diff --git a/.pre-commit-config.yaml b/.pre-commit-config.yaml
index 24bd8fa9..d0740caa 100644
--- a/.pre-commit-config.yaml
+++ b/.pre-commit-config.yaml
@@ -16,6 +16,12 @@ repos:
     hooks:
       - id: black
         args: ["--target-version", "py36"]
+  - repo: https://github.com/pre-commit/mirrors-mypy
+    rev: v0.961
+    hooks:
+      - id: mypy
+        files: ^chardet/
+        args: ["--strict", "--pretty", "--show-error-codes"]
   - repo: https://github.com/PyCQA/prospector
     rev: 1.7.7 # The version of Prospector to use, if not 'master' for latest
     hooks:
diff --git a/MANIFEST.in b/MANIFEST.in
index be5768dd..2f471334 100644
--- a/MANIFEST.in
+++ b/MANIFEST.in
@@ -1,5 +1,6 @@
 include LICENSE
 include *.rst
+include chardet/py.typed
 include requirements.txt
 include test.py
 recursive-include docs *
diff --git a/chardet/__init__.py b/chardet/__init__.py
index e91ad618..2112ee14 100644
--- a/chardet/__init__.py
+++ b/chardet/__init__.py
@@ -15,14 +15,19 @@
 # 02110-1301  USA
 ######################### END LICENSE BLOCK #########################
 
+from typing import List, Union
+
+from .charsetgroupprober import CharSetGroupProber
+from .charsetprober import CharSetProber
 from .enums import InputState
+from .resultdict import ResultDict
 from .universaldetector import UniversalDetector
 from .version import VERSION, __version__
 
 __all__ = ["UniversalDetector", "detect", "detect_all", "__version__", "VERSION"]
 
 
-def detect(byte_str):
+def detect(byte_str: Union[bytes, bytearray]) -> ResultDict:
     """
     Detect the encoding of the given byte string.
 
@@ -40,7 +45,9 @@ def detect(byte_str):
     return detector.close()
 
 
-def detect_all(byte_str, ignore_threshold=False):
+def detect_all(
+    byte_str: Union[bytes, bytearray], ignore_threshold: bool = False
+) -> List[ResultDict]:
     """
     Detect all the possible encodings of the given byte string.
 
@@ -63,10 +70,10 @@ def detect_all(byte_str, ignore_threshold=False):
     detector.close()
 
     if detector.input_state == InputState.HIGH_BYTE:
-        results = []
-        probers = []
+        results: List[ResultDict] = []
+        probers: List[CharSetProber] = []
         for prober in detector.charset_probers:
-            if hasattr(prober, "probers"):
+            if isinstance(prober, CharSetGroupProber):
                 probers.extend(p for p in prober.probers)
             else:
                 probers.append(prober)
diff --git a/chardet/big5prober.py b/chardet/big5prober.py
index e4dfa7aa..ef09c60e 100644
--- a/chardet/big5prober.py
+++ b/chardet/big5prober.py
@@ -32,16 +32,16 @@
 
 
 class Big5Prober(MultiByteCharSetProber):
-    def __init__(self):
+    def __init__(self) -> None:
         super().__init__()
         self.coding_sm = CodingStateMachine(BIG5_SM_MODEL)
         self.distribution_analyzer = Big5DistributionAnalysis()
         self.reset()
 
     @property
-    def charset_name(self):
+    def charset_name(self) -> str:
         return "Big5"
 
     @property
-    def language(self):
+    def language(self) -> str:
         return "Chinese"
diff --git a/chardet/chardistribution.py b/chardet/chardistribution.py
index 27b4a293..176cb996 100644
--- a/chardet/chardistribution.py
+++ b/chardet/chardistribution.py
@@ -25,6 +25,8 @@
 # 02110-1301  USA
 ######################### END LICENSE BLOCK #########################
 
+from typing import Tuple, Union
+
 from .big5freq import (
     BIG5_CHAR_TO_FREQ_ORDER,
     BIG5_TABLE_SIZE,
@@ -59,22 +61,22 @@ class CharDistributionAnalysis:
     SURE_NO = 0.01
     MINIMUM_DATA_THRESHOLD = 3
 
-    def __init__(self):
+    def __init__(self) -> None:
         # Mapping table to get frequency order from char order (get from
         # GetOrder())
-        self._char_to_freq_order = tuple()
-        self._table_size = None  # Size of above table
+        self._char_to_freq_order: Tuple[int, ...] = tuple()
+        self._table_size = 0  # Size of above table
         # This is a constant value which varies from language to language,
         # used in calculating confidence.  See
         # http://www.mozilla.org/projects/intl/UniversalCharsetDetection.html
         # for further detail.
-        self.typical_distribution_ratio = None
-        self._done = None
-        self._total_chars = None
-        self._freq_chars = None
+        self.typical_distribution_ratio = 0.0
+        self._done = False
+        self._total_chars = 0
+        self._freq_chars = 0
         self.reset()
 
-    def reset(self):
+    def reset(self) -> None:
         """reset analyser, clear any state"""
         # If this flag is set to True, detection is done and conclusion has
         # been made
@@ -83,7 +85,7 @@ def reset(self):
         # The number of characters whose frequency order is less than 512
         self._freq_chars = 0
 
-    def feed(self, char, char_len):
+    def feed(self, char: Union[bytes, bytearray], char_len: int) -> None:
         """feed a character with known length"""
         if char_len == 2:
             # we only care about 2-bytes character in our distribution analysis
@@ -97,7 +99,7 @@ def feed(self, char, char_len):
                 if 512 > self._char_to_freq_order[order]:
                     self._freq_chars += 1
 
-    def get_confidence(self):
+    def get_confidence(self) -> float:
         """return confidence based on existing data"""
         # if we didn't receive any character in our consideration range,
         # return negative answer
@@ -114,12 +116,12 @@ def get_confidence(self):
         # normalize confidence (we don't want to be 100% sure)
         return self.SURE_YES
 
-    def got_enough_data(self):
+    def got_enough_data(self) -> bool:
         # It is not necessary to receive all data to draw conclusion.
         # For charset detection, certain amount of data is enough
         return self._total_chars > self.ENOUGH_DATA_THRESHOLD
 
-    def get_order(self, _):
+    def get_order(self, _: Union[bytes, bytearray]) -> int:
         # We do not handle characters based on the original encoding string,
         # but convert this encoding string to a number, here called order.
         # This allows multiple encodings of a language to share one frequency
@@ -128,13 +130,13 @@ def get_order(self, _):
 
 
 class EUCTWDistributionAnalysis(CharDistributionAnalysis):
-    def __init__(self):
+    def __init__(self) -> None:
         super().__init__()
         self._char_to_freq_order = EUCTW_CHAR_TO_FREQ_ORDER
         self._table_size = EUCTW_TABLE_SIZE
         self.typical_distribution_ratio = EUCTW_TYPICAL_DISTRIBUTION_RATIO
 
-    def get_order(self, byte_str):
+    def get_order(self, byte_str: Union[bytes, bytearray]) -> int:
         # for euc-TW encoding, we are interested
         #   first  byte range: 0xc4 -- 0xfe
         #   second byte range: 0xa1 -- 0xfe
@@ -146,13 +148,13 @@ def get_order(self, byte_str):
 
 
 class EUCKRDistributionAnalysis(CharDistributionAnalysis):
-    def __init__(self):
+    def __init__(self) -> None:
         super().__init__()
         self._char_to_freq_order = EUCKR_CHAR_TO_FREQ_ORDER
         self._table_size = EUCKR_TABLE_SIZE
         self.typical_distribution_ratio = EUCKR_TYPICAL_DISTRIBUTION_RATIO
 
-    def get_order(self, byte_str):
+    def get_order(self, byte_str: Union[bytes, bytearray]) -> int:
         # for euc-KR encoding, we are interested
         #   first  byte range: 0xb0 -- 0xfe
         #   second byte range: 0xa1 -- 0xfe
@@ -164,13 +166,13 @@ def get_order(self, byte_str):
 
 
 class JOHABDistributionAnalysis(CharDistributionAnalysis):
-    def __init__(self):
+    def __init__(self) -> None:
         super().__init__()
         self._char_to_freq_order = EUCKR_CHAR_TO_FREQ_ORDER
         self._table_size = EUCKR_TABLE_SIZE
         self.typical_distribution_ratio = EUCKR_TYPICAL_DISTRIBUTION_RATIO
 
-    def get_order(self, byte_str):
+    def get_order(self, byte_str: Union[bytes, bytearray]) -> int:
         first_char = byte_str[0]
         if 0x88 <= first_char < 0xD4:
             code = first_char * 256 + byte_str[1]
@@ -179,13 +181,13 @@ def get_order(self, byte_str):
 
 
 class GB2312DistributionAnalysis(CharDistributionAnalysis):
-    def __init__(self):
+    def __init__(self) -> None:
         super().__init__()
         self._char_to_freq_order = GB2312_CHAR_TO_FREQ_ORDER
         self._table_size = GB2312_TABLE_SIZE
         self.typical_distribution_ratio = GB2312_TYPICAL_DISTRIBUTION_RATIO
 
-    def get_order(self, byte_str):
+    def get_order(self, byte_str: Union[bytes, bytearray]) -> int:
         # for GB2312 encoding, we are interested
         #  first  byte range: 0xb0 -- 0xfe
         #  second byte range: 0xa1 -- 0xfe
@@ -197,13 +199,13 @@ def get_order(self, byte_str):
 
 
 class Big5DistributionAnalysis(CharDistributionAnalysis):
-    def __init__(self):
+    def __init__(self) -> None:
         super().__init__()
         self._char_to_freq_order = BIG5_CHAR_TO_FREQ_ORDER
         self._table_size = BIG5_TABLE_SIZE
         self.typical_distribution_ratio = BIG5_TYPICAL_DISTRIBUTION_RATIO
 
-    def get_order(self, byte_str):
+    def get_order(self, byte_str: Union[bytes, bytearray]) -> int:
         # for big5 encoding, we are interested
         #   first  byte range: 0xa4 -- 0xfe
         #   second byte range: 0x40 -- 0x7e , 0xa1 -- 0xfe
@@ -217,13 +219,13 @@ def get_order(self, byte_str):
 
 
 class SJISDistributionAnalysis(CharDistributionAnalysis):
-    def __init__(self):
+    def __init__(self) -> None:
         super().__init__()
         self._char_to_freq_order = JIS_CHAR_TO_FREQ_ORDER
         self._table_size = JIS_TABLE_SIZE
         self.typical_distribution_ratio = JIS_TYPICAL_DISTRIBUTION_RATIO
 
-    def get_order(self, byte_str):
+    def get_order(self, byte_str: Union[bytes, bytearray]) -> int:
         # for sjis encoding, we are interested
         #   first  byte range: 0x81 -- 0x9f , 0xe0 -- 0xfe
         #   second byte range: 0x40 -- 0x7e,  0x81 -- oxfe
@@ -242,13 +244,13 @@ def get_order(self, byte_str):
 
 
 class EUCJPDistributionAnalysis(CharDistributionAnalysis):
-    def __init__(self):
+    def __init__(self) -> None:
         super().__init__()
         self._char_to_freq_order = JIS_CHAR_TO_FREQ_ORDER
         self._table_size = JIS_TABLE_SIZE
         self.typical_distribution_ratio = JIS_TYPICAL_DISTRIBUTION_RATIO
 
-    def get_order(self, byte_str):
+    def get_order(self, byte_str: Union[bytes, bytearray]) -> int:
         # for euc-JP encoding, we are interested
         #   first  byte range: 0xa0 -- 0xfe
         #   second byte range: 0xa1 -- 0xfe
diff --git a/chardet/charsetgroupprober.py b/chardet/charsetgroupprober.py
index 4c41ed2a..6def56b4 100644
--- a/chardet/charsetgroupprober.py
+++ b/chardet/charsetgroupprober.py
@@ -25,18 +25,20 @@
 # 02110-1301  USA
 ######################### END LICENSE BLOCK #########################
 
+from typing import List, Optional, Union
+
 from .charsetprober import CharSetProber
-from .enums import ProbingState
+from .enums import LanguageFilter, ProbingState
 
 
 class CharSetGroupProber(CharSetProber):
-    def __init__(self, lang_filter=None):
+    def __init__(self, lang_filter: LanguageFilter = LanguageFilter.NONE) -> None:
         super().__init__(lang_filter=lang_filter)
         self._active_num = 0
-        self.probers = []
-        self._best_guess_prober = None
+        self.probers: List[CharSetProber] = []
+        self._best_guess_prober: Optional[CharSetProber] = None
 
-    def reset(self):
+    def reset(self) -> None:
         super().reset()
         self._active_num = 0
         for prober in self.probers:
@@ -46,7 +48,7 @@ def reset(self):
         self._best_guess_prober = None
 
     @property
-    def charset_name(self):
+    def charset_name(self) -> Optional[str]:
         if not self._best_guess_prober:
             self.get_confidence()
             if not self._best_guess_prober:
@@ -54,14 +56,14 @@ def charset_name(self):
         return self._best_guess_prober.charset_name
 
     @property
-    def language(self):
+    def language(self) -> Optional[str]:
         if not self._best_guess_prober:
             self.get_confidence()
             if not self._best_guess_prober:
                 return None
         return self._best_guess_prober.language
 
-    def feed(self, byte_str):
+    def feed(self, byte_str: Union[bytes, bytearray]) -> ProbingState:
         for prober in self.probers:
             if not prober.active:
                 continue
@@ -80,7 +82,7 @@ def feed(self, byte_str):
                     return self.state
         return self.state
 
-    def get_confidence(self):
+    def get_confidence(self) -> float:
         state = self.state
         if state == ProbingState.FOUND_IT:
             return 0.99
diff --git a/chardet/charsetprober.py b/chardet/charsetprober.py
index 9f1afd99..a103ca11 100644
--- a/chardet/charsetprober.py
+++ b/chardet/charsetprober.py
@@ -28,8 +28,9 @@
 
 import logging
 import re
+from typing import Optional, Union
 
-from .enums import ProbingState
+from .enums import LanguageFilter, ProbingState
 
 INTERNATIONAL_WORDS_PATTERN = re.compile(
     b"[a-zA-Z]*[\x80-\xFF]+[a-zA-Z]*[^a-zA-Z\x80-\xFF]?"
@@ -40,35 +41,40 @@ class CharSetProber:
 
     SHORTCUT_THRESHOLD = 0.95
 
-    def __init__(self, lang_filter=None):
-        self._state = None
+    def __init__(self, lang_filter: LanguageFilter = LanguageFilter.NONE) -> None:
+        self._state = ProbingState.DETECTING
+        self.active = True
         self.lang_filter = lang_filter
         self.logger = logging.getLogger(__name__)
 
-    def reset(self):
+    def reset(self) -> None:
         self._state = ProbingState.DETECTING
 
     @property
-    def charset_name(self):
+    def charset_name(self) -> Optional[str]:
         return None
 
-    def feed(self, byte_str):
+    @property
+    def language(self) -> Optional[str]:
+        raise NotImplementedError
+
+    def feed(self, byte_str: Union[bytes, bytearray]) -> ProbingState:
         raise NotImplementedError
 
     @property
-    def state(self):
+    def state(self) -> ProbingState:
         return self._state
 
-    def get_confidence(self):
+    def get_confidence(self) -> float:
         return 0.0
 
     @staticmethod
-    def filter_high_byte_only(buf):
+    def filter_high_byte_only(buf: Union[bytes, bytearray]) -> bytes:
         buf = re.sub(b"([\x00-\x7F])+", b" ", buf)
         return buf
 
     @staticmethod
-    def filter_international_words(buf):
+    def filter_international_words(buf: Union[bytes, bytearray]) -> bytearray:
         """
         We define three types of bytes:
         alphabet: english alphabets [a-zA-Z]
@@ -102,7 +108,7 @@ def filter_international_words(buf):
         return filtered
 
     @staticmethod
-    def remove_xml_tags(buf):
+    def remove_xml_tags(buf: Union[bytes, bytearray]) -> bytes:
         """
         Returns a copy of ``buf`` that retains only the sequences of English
         alphabet and high byte characters that are not between <> characters.
@@ -117,10 +123,13 @@ def remove_xml_tags(buf):
 
         for curr, buf_char in enumerate(buf):
             # Check if we're coming out of or entering an XML tag
-            if buf_char == b">":
+
+            # https://github.com/python/typeshed/issues/8182
+            if buf_char == b">":  # type: ignore[comparison-overlap]
                 prev = curr + 1
                 in_tag = False
-            elif buf_char == b"<":
+            # https://github.com/python/typeshed/issues/8182
+            elif buf_char == b"<":  # type: ignore[comparison-overlap]
                 if curr > prev and not in_tag:
                     # Keep everything after last non-extended-ASCII,
                     # non-alphabetic character
diff --git a/chardet/cli/chardetect.py b/chardet/cli/chardetect.py
index 4959ad6e..94076a4b 100644
--- a/chardet/cli/chardetect.py
+++ b/chardet/cli/chardetect.py
@@ -15,12 +15,15 @@
 
 import argparse
 import sys
+from typing import Iterable, List, Optional
 
 from .. import __version__
 from ..universaldetector import UniversalDetector
 
 
-def description_of(lines, name="stdin", minimal=False):
+def description_of(
+    lines: Iterable[bytes], name: str = "stdin", minimal: bool = False
+) -> Optional[str]:
     """
     Return a string describing the probable encoding of a file or
     list of strings.
@@ -46,7 +49,7 @@ def description_of(lines, name="stdin", minimal=False):
     return f"{name}: no result"
 
 
-def main(argv=None):
+def main(argv: Optional[List[str]] = None) -> None:
     """
     Handles command line arguments and gets things started.
 
diff --git a/chardet/codingstatemachine.py b/chardet/codingstatemachine.py
index d3e3e825..8ed4a877 100644
--- a/chardet/codingstatemachine.py
+++ b/chardet/codingstatemachine.py
@@ -27,6 +27,7 @@
 
 import logging
 
+from .codingstatemachinedict import CodingStateMachineDict
 from .enums import MachineState
 
 
@@ -53,18 +54,19 @@ class CodingStateMachine:
                  encoding from consideration from here on.
     """
 
-    def __init__(self, sm):
+    def __init__(self, sm: CodingStateMachineDict) -> None:
         self._model = sm
         self._curr_byte_pos = 0
         self._curr_char_len = 0
-        self._curr_state = None
+        self._curr_state = MachineState.START
+        self.active = True
         self.logger = logging.getLogger(__name__)
         self.reset()
 
-    def reset(self):
+    def reset(self) -> None:
         self._curr_state = MachineState.START
 
-    def next_state(self, c):
+    def next_state(self, c: int) -> int:
         # for each byte we get its class
         # if it is first byte, we also get byte length
         byte_class = self._model["class_table"][c]
@@ -77,12 +79,12 @@ def next_state(self, c):
         self._curr_byte_pos += 1
         return self._curr_state
 
-    def get_current_charlen(self):
+    def get_current_charlen(self) -> int:
         return self._curr_char_len
 
-    def get_coding_state_machine(self):
+    def get_coding_state_machine(self) -> str:
         return self._model["name"]
 
     @property
-    def language(self):
+    def language(self) -> str:
         return self._model["language"]
diff --git a/chardet/codingstatemachinedict.py b/chardet/codingstatemachinedict.py
new file mode 100644
index 00000000..7a3c4c7e
--- /dev/null
+++ b/chardet/codingstatemachinedict.py
@@ -0,0 +1,19 @@
+from typing import TYPE_CHECKING, Tuple
+
+if TYPE_CHECKING:
+    # TypedDict was introduced in Python 3.8.
+    #
+    # TODO: Remove the else block and TYPE_CHECKING check when dropping support
+    # for Python 3.7.
+    from typing import TypedDict
+
+    class CodingStateMachineDict(TypedDict, total=False):
+        class_table: Tuple[int, ...]
+        class_factor: int
+        state_table: Tuple[int, ...]
+        char_len_table: Tuple[int, ...]
+        name: str
+        language: str  # Optional key
+
+else:
+    CodingStateMachineDict = dict
diff --git a/chardet/cp949prober.py b/chardet/cp949prober.py
index 28a1f3db..fa7307ed 100644
--- a/chardet/cp949prober.py
+++ b/chardet/cp949prober.py
@@ -32,7 +32,7 @@
 
 
 class CP949Prober(MultiByteCharSetProber):
-    def __init__(self):
+    def __init__(self) -> None:
         super().__init__()
         self.coding_sm = CodingStateMachine(CP949_SM_MODEL)
         # NOTE: CP949 is a superset of EUC-KR, so the distribution should be
@@ -41,9 +41,9 @@ def __init__(self):
         self.reset()
 
     @property
-    def charset_name(self):
+    def charset_name(self) -> str:
         return "CP949"
 
     @property
-    def language(self):
+    def language(self) -> str:
         return "Korean"
diff --git a/chardet/enums.py b/chardet/enums.py
index 32a77e76..5e3e1982 100644
--- a/chardet/enums.py
+++ b/chardet/enums.py
@@ -4,6 +4,8 @@
 :author: Dan Blanchard (dan.blanchard@gmail.com)
 """
 
+from enum import Enum, Flag
+
 
 class InputState:
     """
@@ -15,12 +17,13 @@ class InputState:
     HIGH_BYTE = 2
 
 
-class LanguageFilter:
+class LanguageFilter(Flag):
     """
     This enum represents the different language filters we can apply to a
     ``UniversalDetector``.
     """
 
+    NONE = 0x00
     CHINESE_SIMPLIFIED = 0x01
     CHINESE_TRADITIONAL = 0x02
     JAPANESE = 0x04
@@ -31,7 +34,7 @@ class LanguageFilter:
     CJK = CHINESE | JAPANESE | KOREAN
 
 
-class ProbingState:
+class ProbingState(Enum):
     """
     This enum represents the different states a prober can be in.
     """
@@ -62,7 +65,7 @@ class SequenceLikelihood:
     POSITIVE = 3
 
     @classmethod
-    def get_num_categories(cls):
+    def get_num_categories(cls) -> int:
         """:returns: The number of likelihood categories in the enum."""
         return 4
 
diff --git a/chardet/escprober.py b/chardet/escprober.py
index c91c0781..fd713830 100644
--- a/chardet/escprober.py
+++ b/chardet/escprober.py
@@ -25,6 +25,8 @@
 # 02110-1301  USA
 ######################### END LICENSE BLOCK #########################
 
+from typing import Optional, Union
+
 from .charsetprober import CharSetProber
 from .codingstatemachine import CodingStateMachine
 from .enums import LanguageFilter, MachineState, ProbingState
@@ -43,7 +45,7 @@ class EscCharSetProber(CharSetProber):
     identify these encodings.
     """
 
-    def __init__(self, lang_filter=None):
+    def __init__(self, lang_filter: LanguageFilter = LanguageFilter.NONE) -> None:
         super().__init__(lang_filter=lang_filter)
         self.coding_sm = []
         if self.lang_filter & LanguageFilter.CHINESE_SIMPLIFIED:
@@ -53,13 +55,13 @@ def __init__(self, lang_filter=None):
             self.coding_sm.append(CodingStateMachine(ISO2022JP_SM_MODEL))
         if self.lang_filter & LanguageFilter.KOREAN:
             self.coding_sm.append(CodingStateMachine(ISO2022KR_SM_MODEL))
-        self.active_sm_count = None
-        self._detected_charset = None
-        self._detected_language = None
-        self._state = None
+        self.active_sm_count = 0
+        self._detected_charset: Optional[str] = None
+        self._detected_language: Optional[str] = None
+        self._state = ProbingState.DETECTING
         self.reset()
 
-    def reset(self):
+    def reset(self) -> None:
         super().reset()
         for coding_sm in self.coding_sm:
             coding_sm.active = True
@@ -69,17 +71,17 @@ def reset(self):
         self._detected_language = None
 
     @property
-    def charset_name(self):
+    def charset_name(self) -> Optional[str]:
         return self._detected_charset
 
     @property
-    def language(self):
+    def language(self) -> Optional[str]:
         return self._detected_language
 
-    def get_confidence(self):
+    def get_confidence(self) -> float:
         return 0.99 if self._detected_charset else 0.00
 
-    def feed(self, byte_str):
+    def feed(self, byte_str: Union[bytes, bytearray]) -> ProbingState:
         for c in byte_str:
             for coding_sm in self.coding_sm:
                 if not coding_sm.active:
diff --git a/chardet/escsm.py b/chardet/escsm.py
index 3aa0f4d9..11d4adf7 100644
--- a/chardet/escsm.py
+++ b/chardet/escsm.py
@@ -25,6 +25,7 @@
 # 02110-1301  USA
 ######################### END LICENSE BLOCK #########################
 
+from .codingstatemachinedict import CodingStateMachineDict
 from .enums import MachineState
 
 # fmt: off
@@ -75,7 +76,7 @@
 
 HZ_CHAR_LEN_TABLE = (0, 0, 0, 0, 0, 0)
 
-HZ_SM_MODEL = {
+HZ_SM_MODEL: CodingStateMachineDict = {
     "class_table": HZ_CLS,
     "class_factor": 6,
     "state_table": HZ_ST,
@@ -134,7 +135,7 @@
 
 ISO2022CN_CHAR_LEN_TABLE = (0, 0, 0, 0, 0, 0, 0, 0, 0)
 
-ISO2022CN_SM_MODEL = {
+ISO2022CN_SM_MODEL: CodingStateMachineDict = {
     "class_table": ISO2022CN_CLS,
     "class_factor": 9,
     "state_table": ISO2022CN_ST,
@@ -194,7 +195,7 @@
 
 ISO2022JP_CHAR_LEN_TABLE = (0, 0, 0, 0, 0, 0, 0, 0, 0, 0)
 
-ISO2022JP_SM_MODEL = {
+ISO2022JP_SM_MODEL: CodingStateMachineDict = {
     "class_table": ISO2022JP_CLS,
     "class_factor": 10,
     "state_table": ISO2022JP_ST,
@@ -250,7 +251,7 @@
 
 ISO2022KR_CHAR_LEN_TABLE = (0, 0, 0, 0, 0, 0)
 
-ISO2022KR_SM_MODEL = {
+ISO2022KR_SM_MODEL: CodingStateMachineDict = {
     "class_table": ISO2022KR_CLS,
     "class_factor": 6,
     "state_table": ISO2022KR_ST,
diff --git a/chardet/eucjpprober.py b/chardet/eucjpprober.py
index abf2e66e..39487f40 100644
--- a/chardet/eucjpprober.py
+++ b/chardet/eucjpprober.py
@@ -25,6 +25,8 @@
 # 02110-1301  USA
 ######################### END LICENSE BLOCK #########################
 
+from typing import Union
+
 from .chardistribution import EUCJPDistributionAnalysis
 from .codingstatemachine import CodingStateMachine
 from .enums import MachineState, ProbingState
@@ -34,26 +36,29 @@
 
 
 class EUCJPProber(MultiByteCharSetProber):
-    def __init__(self):
+    def __init__(self) -> None:
         super().__init__()
         self.coding_sm = CodingStateMachine(EUCJP_SM_MODEL)
         self.distribution_analyzer = EUCJPDistributionAnalysis()
         self.context_analyzer = EUCJPContextAnalysis()
         self.reset()
 
-    def reset(self):
+    def reset(self) -> None:
         super().reset()
         self.context_analyzer.reset()
 
     @property
-    def charset_name(self):
+    def charset_name(self) -> str:
         return "EUC-JP"
 
     @property
-    def language(self):
+    def language(self) -> str:
         return "Japanese"
 
-    def feed(self, byte_str):
+    def feed(self, byte_str: Union[bytes, bytearray]) -> ProbingState:
+        assert self.coding_sm is not None
+        assert self.distribution_analyzer is not None
+
         for i, byte in enumerate(byte_str):
             # PY3K: byte_str is a byte array, so byte is an int, not a byte
             coding_state = self.coding_sm.next_state(byte)
@@ -89,7 +94,9 @@ def feed(self, byte_str):
 
         return self.state
 
-    def get_confidence(self):
+    def get_confidence(self) -> float:
+        assert self.distribution_analyzer is not None
+
         context_conf = self.context_analyzer.get_confidence()
         distrib_conf = self.distribution_analyzer.get_confidence()
         return max(context_conf, distrib_conf)
diff --git a/chardet/euckrprober.py b/chardet/euckrprober.py
index 154a6d21..1fc5de04 100644
--- a/chardet/euckrprober.py
+++ b/chardet/euckrprober.py
@@ -32,16 +32,16 @@
 
 
 class EUCKRProber(MultiByteCharSetProber):
-    def __init__(self):
+    def __init__(self) -> None:
         super().__init__()
         self.coding_sm = CodingStateMachine(EUCKR_SM_MODEL)
         self.distribution_analyzer = EUCKRDistributionAnalysis()
         self.reset()
 
     @property
-    def charset_name(self):
+    def charset_name(self) -> str:
         return "EUC-KR"
 
     @property
-    def language(self):
+    def language(self) -> str:
         return "Korean"
diff --git a/chardet/euctwprober.py b/chardet/euctwprober.py
index ca10a23c..a37ab189 100644
--- a/chardet/euctwprober.py
+++ b/chardet/euctwprober.py
@@ -32,16 +32,16 @@
 
 
 class EUCTWProber(MultiByteCharSetProber):
-    def __init__(self):
+    def __init__(self) -> None:
         super().__init__()
         self.coding_sm = CodingStateMachine(EUCTW_SM_MODEL)
         self.distribution_analyzer = EUCTWDistributionAnalysis()
         self.reset()
 
     @property
-    def charset_name(self):
+    def charset_name(self) -> str:
         return "EUC-TW"
 
     @property
-    def language(self):
+    def language(self) -> str:
         return "Taiwan"
diff --git a/chardet/gb2312prober.py b/chardet/gb2312prober.py
index 251c0429..d423e731 100644
--- a/chardet/gb2312prober.py
+++ b/chardet/gb2312prober.py
@@ -32,16 +32,16 @@
 
 
 class GB2312Prober(MultiByteCharSetProber):
-    def __init__(self):
+    def __init__(self) -> None:
         super().__init__()
         self.coding_sm = CodingStateMachine(GB2312_SM_MODEL)
         self.distribution_analyzer = GB2312DistributionAnalysis()
         self.reset()
 
     @property
-    def charset_name(self):
+    def charset_name(self) -> str:
         return "GB2312"
 
     @property
-    def language(self):
+    def language(self) -> str:
         return "Chinese"
diff --git a/chardet/hebrewprober.py b/chardet/hebrewprober.py
index 3ca634bf..785d0057 100644
--- a/chardet/hebrewprober.py
+++ b/chardet/hebrewprober.py
@@ -25,8 +25,11 @@
 # 02110-1301  USA
 ######################### END LICENSE BLOCK #########################
 
+from typing import Optional, Union
+
 from .charsetprober import CharSetProber
 from .enums import ProbingState
+from .sbcharsetprober import SingleByteCharSetProber
 
 # This prober doesn't actually recognize a language or a charset.
 # It is a helper prober for the use of the Hebrew model probers
@@ -127,6 +130,7 @@
 
 
 class HebrewProber(CharSetProber):
+    SPACE = 0x20
     # windows-1255 / ISO-8859-8 code points of interest
     FINAL_KAF = 0xEA
     NORMAL_KAF = 0xEB
@@ -152,31 +156,35 @@ class HebrewProber(CharSetProber):
     VISUAL_HEBREW_NAME = "ISO-8859-8"
     LOGICAL_HEBREW_NAME = "windows-1255"
 
-    def __init__(self):
+    def __init__(self) -> None:
         super().__init__()
-        self._final_char_logical_score = None
-        self._final_char_visual_score = None
-        self._prev = None
-        self._before_prev = None
-        self._logical_prober = None
-        self._visual_prober = None
+        self._final_char_logical_score = 0
+        self._final_char_visual_score = 0
+        self._prev = self.SPACE
+        self._before_prev = self.SPACE
+        self._logical_prober: Optional[SingleByteCharSetProber] = None
+        self._visual_prober: Optional[SingleByteCharSetProber] = None
         self.reset()
 
-    def reset(self):
+    def reset(self) -> None:
         self._final_char_logical_score = 0
         self._final_char_visual_score = 0
         # The two last characters seen in the previous buffer,
         # mPrev and mBeforePrev are initialized to space in order to simulate
         # a word delimiter at the beginning of the data
-        self._prev = " "
-        self._before_prev = " "
+        self._prev = self.SPACE
+        self._before_prev = self.SPACE
         # These probers are owned by the group prober.
 
-    def set_model_probers(self, logical_prober, visual_prober):
+    def set_model_probers(
+        self,
+        logical_prober: SingleByteCharSetProber,
+        visual_prober: SingleByteCharSetProber,
+    ) -> None:
         self._logical_prober = logical_prober
         self._visual_prober = visual_prober
 
-    def is_final(self, c):
+    def is_final(self, c: int) -> bool:
         return c in [
             self.FINAL_KAF,
             self.FINAL_MEM,
@@ -185,7 +193,7 @@ def is_final(self, c):
             self.FINAL_TSADI,
         ]
 
-    def is_non_final(self, c):
+    def is_non_final(self, c: int) -> bool:
         # The normal Tsadi is not a good Non-Final letter due to words like
         # 'lechotet' (to chat) containing an apostrophe after the tsadi. This
         # apostrophe is converted to a space in FilterWithoutEnglishLetters
@@ -198,7 +206,7 @@ def is_non_final(self, c):
         # since these words are quite rare.
         return c in [self.NORMAL_KAF, self.NORMAL_MEM, self.NORMAL_NUN, self.NORMAL_PE]
 
-    def feed(self, byte_str):
+    def feed(self, byte_str: Union[bytes, bytearray]) -> ProbingState:
         # Final letter analysis for logical-visual decision.
         # Look for evidence that the received buffer is either logical Hebrew
         # or visual Hebrew.
@@ -232,9 +240,9 @@ def feed(self, byte_str):
         byte_str = self.filter_high_byte_only(byte_str)
 
         for cur in byte_str:
-            if cur == " ":
+            if cur == self.SPACE:
                 # We stand on a space - a word just ended
-                if self._before_prev != " ":
+                if self._before_prev != self.SPACE:
                     # next-to-last char was not a space so self._prev is not a
                     # 1 letter word
                     if self.is_final(self._prev):
@@ -247,9 +255,9 @@ def feed(self, byte_str):
             else:
                 # Not standing on a space
                 if (
-                    (self._before_prev == " ")
+                    (self._before_prev == self.SPACE)
                     and (self.is_final(self._prev))
-                    and (cur != " ")
+                    and (cur != self.SPACE)
                 ):
                     # case (3) [-2:space][-1:final letter][cur:not space]
                     self._final_char_visual_score += 1
@@ -261,7 +269,10 @@ def feed(self, byte_str):
         return ProbingState.DETECTING
 
     @property
-    def charset_name(self):
+    def charset_name(self) -> str:
+        assert self._logical_prober is not None
+        assert self._visual_prober is not None
+
         # Make the decision: is it Logical or Visual?
         # If the final letter score distance is dominant enough, rely on it.
         finalsub = self._final_char_logical_score - self._final_char_visual_score
@@ -289,11 +300,14 @@ def charset_name(self):
         return self.LOGICAL_HEBREW_NAME
 
     @property
-    def language(self):
+    def language(self) -> str:
         return "Hebrew"
 
     @property
-    def state(self):
+    def state(self) -> ProbingState:
+        assert self._logical_prober is not None
+        assert self._visual_prober is not None
+
         # Remain active as long as any of the model probers are active.
         if (self._logical_prober.state == ProbingState.NOT_ME) and (
             self._visual_prober.state == ProbingState.NOT_ME
diff --git a/chardet/johabprober.py b/chardet/johabprober.py
index 6f359d19..d7364ba6 100644
--- a/chardet/johabprober.py
+++ b/chardet/johabprober.py
@@ -32,16 +32,16 @@
 
 
 class JOHABProber(MultiByteCharSetProber):
-    def __init__(self):
+    def __init__(self) -> None:
         super().__init__()
         self.coding_sm = CodingStateMachine(JOHAB_SM_MODEL)
         self.distribution_analyzer = JOHABDistributionAnalysis()
         self.reset()
 
     @property
-    def charset_name(self):
+    def charset_name(self) -> str:
         return "Johab"
 
     @property
-    def language(self):
+    def language(self) -> str:
         return "Korean"
diff --git a/chardet/jpcntx.py b/chardet/jpcntx.py
index 7a8e5be0..2f53bdda 100644
--- a/chardet/jpcntx.py
+++ b/chardet/jpcntx.py
@@ -25,6 +25,7 @@
 # 02110-1301  USA
 ######################### END LICENSE BLOCK #########################
 
+from typing import List, Tuple, Union
 
 # This is hiragana 2-char sequence table, the number in each cell represents its frequency category
 # fmt: off
@@ -123,15 +124,15 @@ class JapaneseContextAnalysis:
     MAX_REL_THRESHOLD = 1000
     MINIMUM_DATA_THRESHOLD = 4
 
-    def __init__(self):
-        self._total_rel = None
-        self._rel_sample = None
-        self._need_to_skip_char_num = None
-        self._last_char_order = None
-        self._done = None
+    def __init__(self) -> None:
+        self._total_rel = 0
+        self._rel_sample: List[int] = []
+        self._need_to_skip_char_num = 0
+        self._last_char_order = -1
+        self._done = False
         self.reset()
 
-    def reset(self):
+    def reset(self) -> None:
         self._total_rel = 0  # total sequence received
         # category counters, each integer counts sequence in its category
         self._rel_sample = [0] * self.NUM_OF_CATEGORY
@@ -143,7 +144,7 @@ def reset(self):
         # been made
         self._done = False
 
-    def feed(self, byte_str, num_bytes):
+    def feed(self, byte_str: Union[bytes, bytearray], num_bytes: int) -> None:
         if self._done:
             return
 
@@ -172,29 +173,29 @@ def feed(self, byte_str, num_bytes):
                     ] += 1
                 self._last_char_order = order
 
-    def got_enough_data(self):
+    def got_enough_data(self) -> bool:
         return self._total_rel > self.ENOUGH_REL_THRESHOLD
 
-    def get_confidence(self):
+    def get_confidence(self) -> float:
         # This is just one way to calculate confidence. It works well for me.
         if self._total_rel > self.MINIMUM_DATA_THRESHOLD:
             return (self._total_rel - self._rel_sample[0]) / self._total_rel
         return self.DONT_KNOW
 
-    def get_order(self, _):
+    def get_order(self, _: Union[bytes, bytearray]) -> Tuple[int, int]:
         return -1, 1
 
 
 class SJISContextAnalysis(JapaneseContextAnalysis):
-    def __init__(self):
+    def __init__(self) -> None:
         super().__init__()
         self._charset_name = "SHIFT_JIS"
 
     @property
-    def charset_name(self):
+    def charset_name(self) -> str:
         return self._charset_name
 
-    def get_order(self, byte_str):
+    def get_order(self, byte_str: Union[bytes, bytearray]) -> Tuple[int, int]:
         if not byte_str:
             return -1, 1
         # find out current char's byte length
@@ -216,7 +217,7 @@ def get_order(self, byte_str):
 
 
 class EUCJPContextAnalysis(JapaneseContextAnalysis):
-    def get_order(self, byte_str):
+    def get_order(self, byte_str: Union[bytes, bytearray]) -> Tuple[int, int]:
         if not byte_str:
             return -1, 1
         # find out current char's byte length
diff --git a/chardet/latin1prober.py b/chardet/latin1prober.py
index 241f14ab..59a01d91 100644
--- a/chardet/latin1prober.py
+++ b/chardet/latin1prober.py
@@ -26,6 +26,8 @@
 # 02110-1301  USA
 ######################### END LICENSE BLOCK #########################
 
+from typing import List, Union
+
 from .charsetprober import CharSetProber
 from .enums import ProbingState
 
@@ -96,26 +98,26 @@
 
 
 class Latin1Prober(CharSetProber):
-    def __init__(self):
+    def __init__(self) -> None:
         super().__init__()
-        self._last_char_class = None
-        self._freq_counter = None
+        self._last_char_class = OTH
+        self._freq_counter: List[int] = []
         self.reset()
 
-    def reset(self):
+    def reset(self) -> None:
         self._last_char_class = OTH
         self._freq_counter = [0] * FREQ_CAT_NUM
         super().reset()
 
     @property
-    def charset_name(self):
+    def charset_name(self) -> str:
         return "ISO-8859-1"
 
     @property
-    def language(self):
+    def language(self) -> str:
         return ""
 
-    def feed(self, byte_str):
+    def feed(self, byte_str: Union[bytes, bytearray]) -> ProbingState:
         byte_str = self.remove_xml_tags(byte_str)
         for c in byte_str:
             char_class = Latin1_CharToClass[c]
@@ -128,7 +130,7 @@ def feed(self, byte_str):
 
         return self.state
 
-    def get_confidence(self):
+    def get_confidence(self) -> float:
         if self.state == ProbingState.NOT_ME:
             return 0.01
 
diff --git a/chardet/macromanprober.py b/chardet/macromanprober.py
index a7bf5b41..1425d10e 100644
--- a/chardet/macromanprober.py
+++ b/chardet/macromanprober.py
@@ -28,6 +28,8 @@
 # 02110-1301  USA
 ######################### END LICENSE BLOCK #########################
 
+from typing import List, Union
+
 from .charsetprober import CharSetProber
 from .enums import ProbingState
 
@@ -105,13 +107,13 @@
 
 
 class MacRomanProber(CharSetProber):
-    def __init__(self):
+    def __init__(self) -> None:
         super().__init__()
-        self._last_char_class = None
-        self._freq_counter = None
+        self._last_char_class = OTH
+        self._freq_counter: List[int] = []
         self.reset()
 
-    def reset(self):
+    def reset(self) -> None:
         self._last_char_class = OTH
         self._freq_counter = [0] * FREQ_CAT_NUM
 
@@ -123,14 +125,14 @@ def reset(self):
         super().reset()
 
     @property
-    def charset_name(self):
+    def charset_name(self) -> str:
         return "MacRoman"
 
     @property
-    def language(self):
+    def language(self) -> str:
         return ""
 
-    def feed(self, byte_str):
+    def feed(self, byte_str: Union[bytes, bytearray]) -> ProbingState:
         byte_str = self.remove_xml_tags(byte_str)
         for c in byte_str:
             char_class = MacRoman_CharToClass[c]
@@ -143,7 +145,7 @@ def feed(self, byte_str):
 
         return self.state
 
-    def get_confidence(self):
+    def get_confidence(self) -> float:
         if self.state == ProbingState.NOT_ME:
             return 0.01
 
diff --git a/chardet/mbcharsetprober.py b/chardet/mbcharsetprober.py
index bf96ad5d..666307e8 100644
--- a/chardet/mbcharsetprober.py
+++ b/chardet/mbcharsetprober.py
@@ -27,8 +27,12 @@
 # 02110-1301  USA
 ######################### END LICENSE BLOCK #########################
 
+from typing import Optional, Union
+
+from .chardistribution import CharDistributionAnalysis
 from .charsetprober import CharSetProber
-from .enums import MachineState, ProbingState
+from .codingstatemachine import CodingStateMachine
+from .enums import LanguageFilter, MachineState, ProbingState
 
 
 class MultiByteCharSetProber(CharSetProber):
@@ -36,29 +40,24 @@ class MultiByteCharSetProber(CharSetProber):
     MultiByteCharSetProber
     """
 
-    def __init__(self, lang_filter=None):
+    def __init__(self, lang_filter: LanguageFilter = LanguageFilter.NONE) -> None:
         super().__init__(lang_filter=lang_filter)
-        self.distribution_analyzer = None
-        self.coding_sm = None
-        self._last_char = [0, 0]
+        self.distribution_analyzer: Optional[CharDistributionAnalysis] = None
+        self.coding_sm: Optional[CodingStateMachine] = None
+        self._last_char = bytearray(b"\0\0")
 
-    def reset(self):
+    def reset(self) -> None:
         super().reset()
         if self.coding_sm:
             self.coding_sm.reset()
         if self.distribution_analyzer:
             self.distribution_analyzer.reset()
-        self._last_char = [0, 0]
-
-    @property
-    def charset_name(self):
-        raise NotImplementedError
+        self._last_char = bytearray(b"\0\0")
 
-    @property
-    def language(self):
-        raise NotImplementedError
+    def feed(self, byte_str: Union[bytes, bytearray]) -> ProbingState:
+        assert self.coding_sm is not None
+        assert self.distribution_analyzer is not None
 
-    def feed(self, byte_str):
         for i, byte in enumerate(byte_str):
             coding_state = self.coding_sm.next_state(byte)
             if coding_state == MachineState.ERROR:
@@ -91,5 +90,6 @@ def feed(self, byte_str):
 
         return self.state
 
-    def get_confidence(self):
+    def get_confidence(self) -> float:
+        assert self.distribution_analyzer is not None
         return self.distribution_analyzer.get_confidence()
diff --git a/chardet/mbcsgroupprober.py b/chardet/mbcsgroupprober.py
index 94488360..6cb9cc7b 100644
--- a/chardet/mbcsgroupprober.py
+++ b/chardet/mbcsgroupprober.py
@@ -30,6 +30,7 @@
 from .big5prober import Big5Prober
 from .charsetgroupprober import CharSetGroupProber
 from .cp949prober import CP949Prober
+from .enums import LanguageFilter
 from .eucjpprober import EUCJPProber
 from .euckrprober import EUCKRProber
 from .euctwprober import EUCTWProber
@@ -40,7 +41,7 @@
 
 
 class MBCSGroupProber(CharSetGroupProber):
-    def __init__(self, lang_filter=None):
+    def __init__(self, lang_filter: LanguageFilter = LanguageFilter.NONE) -> None:
         super().__init__(lang_filter=lang_filter)
         self.probers = [
             UTF8Prober(),
diff --git a/chardet/mbcssm.py b/chardet/mbcssm.py
index d3b9c4b7..7bbe97e6 100644
--- a/chardet/mbcssm.py
+++ b/chardet/mbcssm.py
@@ -25,6 +25,7 @@
 # 02110-1301  USA
 ######################### END LICENSE BLOCK #########################
 
+from .codingstatemachinedict import CodingStateMachineDict
 from .enums import MachineState
 
 # BIG5
@@ -74,7 +75,7 @@
 
 BIG5_CHAR_LEN_TABLE = (0, 1, 1, 2, 0)
 
-BIG5_SM_MODEL = {
+BIG5_SM_MODEL: CodingStateMachineDict = {
     "class_table": BIG5_CLS,
     "class_factor": 5,
     "state_table": BIG5_ST,
@@ -117,7 +118,7 @@
 
 CP949_CHAR_LEN_TABLE = (0, 1, 2, 0, 1, 1, 2, 2, 0, 2)
 
-CP949_SM_MODEL = {
+CP949_SM_MODEL: CodingStateMachineDict = {
     "class_table": CP949_CLS,
     "class_factor": 10,
     "state_table": CP949_ST,
@@ -173,7 +174,7 @@
 
 EUCJP_CHAR_LEN_TABLE = (2, 2, 2, 3, 1, 0)
 
-EUCJP_SM_MODEL = {
+EUCJP_SM_MODEL: CodingStateMachineDict = {
     "class_table": EUCJP_CLS,
     "class_factor": 6,
     "state_table": EUCJP_ST,
@@ -226,7 +227,7 @@
 
 EUCKR_CHAR_LEN_TABLE = (0, 1, 2, 0)
 
-EUCKR_SM_MODEL = {
+EUCKR_SM_MODEL: CodingStateMachineDict = {
     "class_table": EUCKR_CLS,
     "class_factor": 4,
     "state_table": EUCKR_ST,
@@ -283,7 +284,7 @@
 
 JOHAB_CHAR_LEN_TABLE = (0, 1, 1, 1, 1, 0, 0, 2, 2, 2)
 
-JOHAB_SM_MODEL = {
+JOHAB_SM_MODEL: CodingStateMachineDict = {
     "class_table": JOHAB_CLS,
     "class_factor": 10,
     "state_table": JOHAB_ST,
@@ -340,7 +341,7 @@
 
 EUCTW_CHAR_LEN_TABLE = (0, 0, 1, 2, 2, 2, 3)
 
-EUCTW_SM_MODEL = {
+EUCTW_SM_MODEL: CodingStateMachineDict = {
     "class_table": EUCTW_CLS,
     "class_factor": 7,
     "state_table": EUCTW_ST,
@@ -402,7 +403,7 @@
 # 2 here.
 GB2312_CHAR_LEN_TABLE = (0, 1, 1, 1, 1, 1, 2)
 
-GB2312_SM_MODEL = {
+GB2312_SM_MODEL: CodingStateMachineDict = {
     "class_table": GB2312_CLS,
     "class_factor": 7,
     "state_table": GB2312_ST,
@@ -458,7 +459,7 @@
 
 SJIS_CHAR_LEN_TABLE = (0, 1, 1, 2, 0, 0)
 
-SJIS_SM_MODEL = {
+SJIS_SM_MODEL: CodingStateMachineDict = {
     "class_table": SJIS_CLS,
     "class_factor": 6,
     "state_table": SJIS_ST,
@@ -516,7 +517,7 @@
 
 UCS2BE_CHAR_LEN_TABLE = (2, 2, 2, 0, 2, 2)
 
-UCS2BE_SM_MODEL = {
+UCS2BE_SM_MODEL: CodingStateMachineDict = {
     "class_table": UCS2BE_CLS,
     "class_factor": 6,
     "state_table": UCS2BE_ST,
@@ -574,7 +575,7 @@
 
 UCS2LE_CHAR_LEN_TABLE = (2, 2, 2, 2, 2, 2)
 
-UCS2LE_SM_MODEL = {
+UCS2LE_SM_MODEL: CodingStateMachineDict = {
     "class_table": UCS2LE_CLS,
     "class_factor": 6,
     "state_table": UCS2LE_ST,
@@ -651,7 +652,7 @@
 
 UTF8_CHAR_LEN_TABLE = (0, 1, 0, 0, 0, 0, 2, 3, 3, 3, 4, 4, 5, 5, 6, 6)
 
-UTF8_SM_MODEL = {
+UTF8_SM_MODEL: CodingStateMachineDict = {
     "class_table": UTF8_CLS,
     "class_factor": 16,
     "state_table": UTF8_ST,
diff --git a/chardet/metadata/languages.py b/chardet/metadata/languages.py
index e06ec9f5..558a8981 100644
--- a/chardet/metadata/languages.py
+++ b/chardet/metadata/languages.py
@@ -6,6 +6,7 @@
 """
 
 from string import ascii_letters
+from typing import List, Optional
 
 # TODO: Add Ukrainian (KOI8-U)
 
@@ -33,13 +34,13 @@ class Language:
 
     def __init__(
         self,
-        name=None,
-        iso_code=None,
-        use_ascii=True,
-        charsets=None,
-        alphabet=None,
-        wiki_start_pages=None,
-    ):
+        name: Optional[str] = None,
+        iso_code: Optional[str] = None,
+        use_ascii: bool = True,
+        charsets: Optional[List[str]] = None,
+        alphabet: Optional[str] = None,
+        wiki_start_pages: Optional[List[str]] = None,
+    ) -> None:
         super().__init__()
         self.name = name
         self.iso_code = iso_code
@@ -55,7 +56,7 @@ def __init__(
         self.alphabet = "".join(sorted(set(alphabet))) if alphabet else None
         self.wiki_start_pages = wiki_start_pages
 
-    def __repr__(self):
+    def __repr__(self) -> str:
         param_str = ", ".join(
             f"{k}={v!r}" for k, v in self.__dict__.items() if not k.startswith("_")
         )
diff --git a/chardet/py.typed b/chardet/py.typed
new file mode 100644
index 00000000..e69de29b
diff --git a/chardet/resultdict.py b/chardet/resultdict.py
new file mode 100644
index 00000000..7d36e64c
--- /dev/null
+++ b/chardet/resultdict.py
@@ -0,0 +1,16 @@
+from typing import TYPE_CHECKING, Optional
+
+if TYPE_CHECKING:
+    # TypedDict was introduced in Python 3.8.
+    #
+    # TODO: Remove the else block and TYPE_CHECKING check when dropping support
+    # for Python 3.7.
+    from typing import TypedDict
+
+    class ResultDict(TypedDict):
+        encoding: Optional[str]
+        confidence: float
+        language: Optional[str]
+
+else:
+    ResultDict = dict
diff --git a/chardet/sbcharsetprober.py b/chardet/sbcharsetprober.py
index 31d70e15..0ffbcdd2 100644
--- a/chardet/sbcharsetprober.py
+++ b/chardet/sbcharsetprober.py
@@ -26,23 +26,20 @@
 # 02110-1301  USA
 ######################### END LICENSE BLOCK #########################
 
-from collections import namedtuple
+from typing import Dict, List, NamedTuple, Optional, Union
 
 from .charsetprober import CharSetProber
 from .enums import CharacterCategory, ProbingState, SequenceLikelihood
 
-SingleByteCharSetModel = namedtuple(
-    "SingleByteCharSetModel",
-    [
-        "charset_name",
-        "language",
-        "char_to_order_map",
-        "language_model",
-        "typical_positive_ratio",
-        "keep_ascii_letters",
-        "alphabet",
-    ],
-)
+
+class SingleByteCharSetModel(NamedTuple):
+    charset_name: str
+    language: str
+    char_to_order_map: Dict[int, int]
+    language_model: Dict[int, Dict[int, int]]
+    typical_positive_ratio: float
+    keep_ascii_letters: bool
+    alphabet: str
 
 
 class SingleByteCharSetProber(CharSetProber):
@@ -51,22 +48,27 @@ class SingleByteCharSetProber(CharSetProber):
     POSITIVE_SHORTCUT_THRESHOLD = 0.95
     NEGATIVE_SHORTCUT_THRESHOLD = 0.05
 
-    def __init__(self, model, is_reversed=False, name_prober=None):
+    def __init__(
+        self,
+        model: SingleByteCharSetModel,
+        is_reversed: bool = False,
+        name_prober: Optional[CharSetProber] = None,
+    ) -> None:
         super().__init__()
         self._model = model
         # TRUE if we need to reverse every pair in the model lookup
         self._reversed = is_reversed
         # Optional auxiliary prober for name decision
         self._name_prober = name_prober
-        self._last_order = None
-        self._seq_counters = None
-        self._total_seqs = None
-        self._total_char = None
-        self._control_char = None
-        self._freq_char = None
+        self._last_order = 255
+        self._seq_counters: List[int] = []
+        self._total_seqs = 0
+        self._total_char = 0
+        self._control_char = 0
+        self._freq_char = 0
         self.reset()
 
-    def reset(self):
+    def reset(self) -> None:
         super().reset()
         # char order of last character
         self._last_order = 255
@@ -78,18 +80,18 @@ def reset(self):
         self._freq_char = 0
 
     @property
-    def charset_name(self):
+    def charset_name(self) -> Optional[str]:
         if self._name_prober:
             return self._name_prober.charset_name
         return self._model.charset_name
 
     @property
-    def language(self):
+    def language(self) -> Optional[str]:
         if self._name_prober:
             return self._name_prober.language
         return self._model.language
 
-    def feed(self, byte_str):
+    def feed(self, byte_str: Union[bytes, bytearray]) -> ProbingState:
         # TODO: Make filter_international_words keep things in self.alphabet
         if not self._model.keep_ascii_letters:
             byte_str = self.filter_international_words(byte_str)
@@ -139,7 +141,7 @@ def feed(self, byte_str):
 
         return self.state
 
-    def get_confidence(self):
+    def get_confidence(self) -> float:
         r = 0.01
         if self._total_seqs > 0:
             r = (
diff --git a/chardet/sbcsgroupprober.py b/chardet/sbcsgroupprober.py
index cad001cb..890ae846 100644
--- a/chardet/sbcsgroupprober.py
+++ b/chardet/sbcsgroupprober.py
@@ -48,7 +48,7 @@
 
 
 class SBCSGroupProber(CharSetGroupProber):
-    def __init__(self):
+    def __init__(self) -> None:
         super().__init__()
         hebrew_prober = HebrewProber()
         logical_hebrew_prober = SingleByteCharSetProber(
diff --git a/chardet/sjisprober.py b/chardet/sjisprober.py
index 3bcbdb71..91df0779 100644
--- a/chardet/sjisprober.py
+++ b/chardet/sjisprober.py
@@ -25,6 +25,8 @@
 # 02110-1301  USA
 ######################### END LICENSE BLOCK #########################
 
+from typing import Union
+
 from .chardistribution import SJISDistributionAnalysis
 from .codingstatemachine import CodingStateMachine
 from .enums import MachineState, ProbingState
@@ -34,26 +36,29 @@
 
 
 class SJISProber(MultiByteCharSetProber):
-    def __init__(self):
+    def __init__(self) -> None:
         super().__init__()
         self.coding_sm = CodingStateMachine(SJIS_SM_MODEL)
         self.distribution_analyzer = SJISDistributionAnalysis()
         self.context_analyzer = SJISContextAnalysis()
         self.reset()
 
-    def reset(self):
+    def reset(self) -> None:
         super().reset()
         self.context_analyzer.reset()
 
     @property
-    def charset_name(self):
+    def charset_name(self) -> str:
         return self.context_analyzer.charset_name
 
     @property
-    def language(self):
+    def language(self) -> str:
         return "Japanese"
 
-    def feed(self, byte_str):
+    def feed(self, byte_str: Union[bytes, bytearray]) -> ProbingState:
+        assert self.coding_sm is not None
+        assert self.distribution_analyzer is not None
+
         for i, byte in enumerate(byte_str):
             coding_state = self.coding_sm.next_state(byte)
             if coding_state == MachineState.ERROR:
@@ -92,7 +97,9 @@ def feed(self, byte_str):
 
         return self.state
 
-    def get_confidence(self):
+    def get_confidence(self) -> float:
+        assert self.distribution_analyzer is not None
+
         context_conf = self.context_analyzer.get_confidence()
         distrib_conf = self.distribution_analyzer.get_confidence()
         return max(context_conf, distrib_conf)
diff --git a/chardet/universaldetector.py b/chardet/universaldetector.py
index de2d4c0f..f94961e6 100644
--- a/chardet/universaldetector.py
+++ b/chardet/universaldetector.py
@@ -39,13 +39,16 @@ class a user of ``chardet`` should use.
 import codecs
 import logging
 import re
+from typing import List, Optional, Union
 
 from .charsetgroupprober import CharSetGroupProber
+from .charsetprober import CharSetProber
 from .enums import InputState, LanguageFilter, ProbingState
 from .escprober import EscCharSetProber
 from .latin1prober import Latin1Prober
 from .macromanprober import MacRomanProber
 from .mbcsgroupprober import MBCSGroupProber
+from .resultdict import ResultDict
 from .sbcsgroupprober import SBCSGroupProber
 from .utf1632prober import UTF1632Prober
 
@@ -82,33 +85,37 @@ class UniversalDetector:
         "iso-8859-13": "Windows-1257",
     }
 
-    def __init__(self, lang_filter=LanguageFilter.ALL):
-        self._esc_charset_prober = None
-        self._utf1632_prober = None
-        self._charset_probers = []
-        self.result = None
-        self.done = None
-        self._got_data = None
-        self._input_state = None
-        self._last_char = None
+    def __init__(self, lang_filter: LanguageFilter = LanguageFilter.ALL) -> None:
+        self._esc_charset_prober: Optional[EscCharSetProber] = None
+        self._utf1632_prober: Optional[UTF1632Prober] = None
+        self._charset_probers: List[CharSetProber] = []
+        self.result: ResultDict = {
+            "encoding": None,
+            "confidence": 0.0,
+            "language": None,
+        }
+        self.done = False
+        self._got_data = False
+        self._input_state = InputState.PURE_ASCII
+        self._last_char = b""
         self.lang_filter = lang_filter
         self.logger = logging.getLogger(__name__)
-        self._has_win_bytes = None
+        self._has_win_bytes = False
         self.reset()
 
     @property
-    def input_state(self):
+    def input_state(self) -> int:
         return self._input_state
 
     @property
-    def has_win_bytes(self):
+    def has_win_bytes(self) -> bool:
         return self._has_win_bytes
 
     @property
-    def charset_probers(self):
+    def charset_probers(self) -> List[CharSetProber]:
         return self._charset_probers
 
-    def reset(self):
+    def reset(self) -> None:
         """
         Reset the UniversalDetector and all of its probers back to their
         initial states.  This is called by ``__init__``, so you only need to
@@ -127,7 +134,7 @@ def reset(self):
         for prober in self._charset_probers:
             prober.reset()
 
-    def feed(self, byte_str):
+    def feed(self, byte_str: Union[bytes, bytearray]) -> None:
         """
         Takes a chunk of a document and feeds it through all of the relevant
         charset probers.
@@ -256,7 +263,7 @@ def feed(self, byte_str):
             if self.WIN_BYTE_DETECTOR.search(byte_str):
                 self._has_win_bytes = True
 
-    def close(self):
+    def close(self) -> ResultDict:
         """
         Stop analyzing the current document and come up with a final
         prediction.
@@ -290,7 +297,8 @@ def close(self):
                     max_prober = prober
             if max_prober and (max_prober_confidence > self.MINIMUM_THRESHOLD):
                 charset_name = max_prober.charset_name
-                lower_charset_name = max_prober.charset_name.lower()
+                assert charset_name is not None
+                lower_charset_name = charset_name.lower()
                 confidence = max_prober.get_confidence()
                 # Use Windows encoding name instead of ISO-8859 if we saw any
                 # extra Windows-specific bytes
diff --git a/chardet/utf1632prober.py b/chardet/utf1632prober.py
index 9fd1580b..6bdec63d 100644
--- a/chardet/utf1632prober.py
+++ b/chardet/utf1632prober.py
@@ -18,6 +18,8 @@
 # Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA
 # 02110-1301  USA
 ######################### END LICENSE BLOCK #########################
+from typing import List, Union
+
 from .charsetprober import CharSetProber
 from .enums import ProbingState
 
@@ -36,7 +38,7 @@ class UTF1632Prober(CharSetProber):
     # a fixed constant ratio of expected zeros or non-zeros in modulo-position.
     EXPECTED_RATIO = 0.94
 
-    def __init__(self):
+    def __init__(self) -> None:
         super().__init__()
         self.position = 0
         self.zeros_at_mod = [0] * 4
@@ -51,7 +53,7 @@ def __init__(self):
         self.first_half_surrogate_pair_detected_16le = False
         self.reset()
 
-    def reset(self):
+    def reset(self) -> None:
         super().reset()
         self.position = 0
         self.zeros_at_mod = [0] * 4
@@ -66,7 +68,7 @@ def reset(self):
         self.quad = [0, 0, 0, 0]
 
     @property
-    def charset_name(self):
+    def charset_name(self) -> str:
         if self.is_likely_utf32be():
             return "utf-32be"
         if self.is_likely_utf32le():
@@ -79,16 +81,16 @@ def charset_name(self):
         return "utf-16"
 
     @property
-    def language(self):
+    def language(self) -> str:
         return ""
 
-    def approx_32bit_chars(self):
+    def approx_32bit_chars(self) -> float:
         return max(1.0, self.position / 4.0)
 
-    def approx_16bit_chars(self):
+    def approx_16bit_chars(self) -> float:
         return max(1.0, self.position / 2.0)
 
-    def is_likely_utf32be(self):
+    def is_likely_utf32be(self) -> bool:
         approx_chars = self.approx_32bit_chars()
         return approx_chars >= self.MIN_CHARS_FOR_DETECTION and (
             self.zeros_at_mod[0] / approx_chars > self.EXPECTED_RATIO
@@ -98,7 +100,7 @@ def is_likely_utf32be(self):
             and not self.invalid_utf32be
         )
 
-    def is_likely_utf32le(self):
+    def is_likely_utf32le(self) -> bool:
         approx_chars = self.approx_32bit_chars()
         return approx_chars >= self.MIN_CHARS_FOR_DETECTION and (
             self.nonzeros_at_mod[0] / approx_chars > self.EXPECTED_RATIO
@@ -108,7 +110,7 @@ def is_likely_utf32le(self):
             and not self.invalid_utf32le
         )
 
-    def is_likely_utf16be(self):
+    def is_likely_utf16be(self) -> bool:
         approx_chars = self.approx_16bit_chars()
         return approx_chars >= self.MIN_CHARS_FOR_DETECTION and (
             (self.nonzeros_at_mod[1] + self.nonzeros_at_mod[3]) / approx_chars
@@ -118,7 +120,7 @@ def is_likely_utf16be(self):
             and not self.invalid_utf16be
         )
 
-    def is_likely_utf16le(self):
+    def is_likely_utf16le(self) -> bool:
         approx_chars = self.approx_16bit_chars()
         return approx_chars >= self.MIN_CHARS_FOR_DETECTION and (
             (self.nonzeros_at_mod[0] + self.nonzeros_at_mod[2]) / approx_chars
@@ -128,7 +130,7 @@ def is_likely_utf16le(self):
             and not self.invalid_utf16le
         )
 
-    def validate_utf32_characters(self, quad):
+    def validate_utf32_characters(self, quad: List[int]) -> None:
         """
         Validate if the quad of bytes is valid UTF-32.
 
@@ -150,7 +152,7 @@ def validate_utf32_characters(self, quad):
         ):
             self.invalid_utf32le = True
 
-    def validate_utf16_characters(self, pair):
+    def validate_utf16_characters(self, pair: List[int]) -> None:
         """
         Validate if the pair of bytes is  valid UTF-16.
 
@@ -182,7 +184,7 @@ def validate_utf16_characters(self, pair):
             else:
                 self.invalid_utf16le = True
 
-    def feed(self, byte_str):
+    def feed(self, byte_str: Union[bytes, bytearray]) -> ProbingState:
         for c in byte_str:
             mod4 = self.position % 4
             self.quad[mod4] = c
@@ -198,7 +200,7 @@ def feed(self, byte_str):
         return self.state
 
     @property
-    def state(self):
+    def state(self) -> ProbingState:
         if self._state in {ProbingState.NOT_ME, ProbingState.FOUND_IT}:
             # terminal, decided states
             return self._state
@@ -210,7 +212,7 @@ def state(self):
             self._state = ProbingState.NOT_ME
         return self._state
 
-    def get_confidence(self):
+    def get_confidence(self) -> float:
         return (
             0.85
             if (
diff --git a/chardet/utf8prober.py b/chardet/utf8prober.py
index 3aae09e8..d96354d9 100644
--- a/chardet/utf8prober.py
+++ b/chardet/utf8prober.py
@@ -25,6 +25,8 @@
 # 02110-1301  USA
 ######################### END LICENSE BLOCK #########################
 
+from typing import Union
+
 from .charsetprober import CharSetProber
 from .codingstatemachine import CodingStateMachine
 from .enums import MachineState, ProbingState
@@ -34,26 +36,26 @@
 class UTF8Prober(CharSetProber):
     ONE_CHAR_PROB = 0.5
 
-    def __init__(self):
+    def __init__(self) -> None:
         super().__init__()
         self.coding_sm = CodingStateMachine(UTF8_SM_MODEL)
-        self._num_mb_chars = None
+        self._num_mb_chars = 0
         self.reset()
 
-    def reset(self):
+    def reset(self) -> None:
         super().reset()
         self.coding_sm.reset()
         self._num_mb_chars = 0
 
     @property
-    def charset_name(self):
+    def charset_name(self) -> str:
         return "utf-8"
 
     @property
-    def language(self):
+    def language(self) -> str:
         return ""
 
-    def feed(self, byte_str):
+    def feed(self, byte_str: Union[bytes, bytearray]) -> ProbingState:
         for c in byte_str:
             coding_state = self.coding_sm.next_state(c)
             if coding_state == MachineState.ERROR:
@@ -72,7 +74,7 @@ def feed(self, byte_str):
 
         return self.state
 
-    def get_confidence(self):
+    def get_confidence(self) -> float:
         unlike = 0.99
         if self._num_mb_chars < 6:
             unlike *= self.ONE_CHAR_PROB**self._num_mb_chars
diff --git a/setup.cfg b/setup.cfg
index f13efeeb..1a4a9c65 100644
--- a/setup.cfg
+++ b/setup.cfg
@@ -31,9 +31,13 @@ classifiers =
     Topic :: Text Processing :: Linguistic
 
 [options]
+include_package_data = true
 python_requires = >=3.7
 packages = find:
 
+[options.package_data]
+chardet = py.typed
+
 [options.entry_points]
 console_scripts =
     chardetect = chardet.cli.chardetect:main