Fast, correct Python JSON library supporting dataclasses, datetimes, and numpy



orjson is a fast, correct JSON library for Python. It benchmarks as the fastest Python library for JSON and is more correct than the standard json library or other third-party libraries. It serializes dataclass, datetime, numpy, and UUID instances natively.

Its features and drawbacks compared to other Python JSON libraries:

  • serializes dataclass instances 40-50x as fast as other libraries
  • serializes datetime, date, and time instances to RFC 3339 format, e.g., "1970-01-01T00:00:00+00:00"
  • serializes numpy.ndarray instances 4-12x as fast with 0.3x the memory usage of other libraries
  • pretty prints 10x to 20x as fast as the standard library
  • serializes to bytes rather than str, i.e., is not a drop-in replacement
  • serializes str without escaping unicode to ASCII, e.g., "好" rather than "\\u597d"
  • serializes float 10x as fast and deserializes twice as fast as other libraries
  • serializes subclasses of str, int, list, and dict natively, requiring default to specify how to serialize others
  • serializes arbitrary types using a default hook
  • has strict UTF-8 conformance, more correct than the standard library
  • has strict JSON conformance in not supporting Nan/Infinity/-Infinity
  • has an option for strict JSON conformance on 53-bit integers with default support for 64-bit
  • does not provide load() or dump() functions for reading from/writing to file-like objects

orjson supports CPython 3.6, 3.7, 3.8, 3.9, and 3.10. It distributes x86_64/amd64 and aarch64/armv8 wheels for Linux. It distributes x86_64/amd64 wheels for macOS and Windows. orjson does not support PyPy. Releases follow semantic versioning and serializing a new object type without an opt-in flag is considered a breaking change.

orjson is licensed under both the Apache 2.0 and MIT licenses. The repository and issue tracker is, and patches may be submitted there. There is a CHANGELOG available in the repository.

  1. Usage
    1. Install
    2. Quickstart
    3. Migrating
    4. Serialize
      1. default
      2. option
    5. Deserialize
  2. Types
    1. dataclass
    2. datetime
    3. enum
    4. float
    5. int
    6. numpy
    7. str
    8. uuid
  3. Testing
  4. Performance
    1. Latency
    2. Memory
    3. Reproducing
  5. Questions
  6. Packaging
  7. License



To install a wheel from PyPI:

pip install --upgrade "pip>=19.3" # manylinux2014 support
pip install --upgrade orjson

Notice that Linux environments with a pip version shipped in 2018 or earlier must first upgrade pip to support manylinux2014 wheels.

To build a wheel, see packaging.


This is an example of serializing, with options specified, and deserializing:

>>> import orjson, datetime, numpy
>>> data = {
    "type": "job",
    "created_at": datetime.datetime(1970, 1, 1),
    "status": "🆗",
    "payload": numpy.array([[1, 2], [3, 4]]),
>>> orjson.dumps(data, option=orjson.OPT_NAIVE_UTC | orjson.OPT_SERIALIZE_NUMPY)
>>> orjson.loads(_)
{'type': 'job', 'created_at': '1970-01-01T00:00:00+00:00', 'status': '🆗', 'payload': [[1, 2], [3, 4]]}


orjson version 3 serializes more types than version 2. Subclasses of str, int, dict, and list are now serialized. This is faster and more similar to the standard library. It can be disabled with orjson.OPT_PASSTHROUGH_SUBCLASS.dataclasses.dataclass instances are now serialized by default and cannot be customized in a default function unless option=orjson.OPT_PASSTHROUGH_DATACLASS is specified. uuid.UUID instances are serialized by default. For any type that is now serialized, implementations in a default function and options enabling them can be removed but do not need to be. There was no change in deserialization.

To migrate from the standard library, the largest difference is that orjson.dumps returns bytes and json.dumps returns a str. Users with dict objects using non-str keys should specify option=orjson.OPT_NON_STR_KEYS. sort_keys is replaced by option=orjson.OPT_SORT_KEYS. indent is replaced by option=orjson.OPT_INDENT_2 and other levels of indentation are not supported.


def dumps(
    __obj: Any,
    default: Optional[Callable[[Any], Any]] = ...,
    option: Optional[int] = ...,
) -> bytes: ...

dumps() serializes Python objects to JSON.

It natively serializes str, dict, list, tuple, int, float, bool, dataclasses.dataclass, typing.TypedDict, datetime.datetime,, datetime.time, uuid.UUID, numpy.ndarray, and None instances. It supports arbitrary types through default. It serializes subclasses of str, int, dict, list, dataclasses.dataclass, and enum.Enum. It does not serialize subclasses of tuple to avoid serializing namedtuple objects as arrays. To avoid serializing subclasses, specify the option orjson.OPT_PASSTHROUGH_SUBCLASS.

The output is a bytes object containing UTF-8.

The global interpreter lock (GIL) is held for the duration of the call.

It raises JSONEncodeError on an unsupported type. This exception message describes the invalid object with the error message Type is not JSON serializable: .... To fix this, specify default.

It raises JSONEncodeError on a str that contains invalid UTF-8.

It raises JSONEncodeError on an integer that exceeds 64 bits by default or, with OPT_STRICT_INTEGER, 53 bits.

It raises JSONEncodeError if a dict has a key of a type other than str, unless OPT_NON_STR_KEYS is specified.

It raises JSONEncodeError if the output of default recurses to handling by default more than 254 levels deep.

It raises JSONEncodeError on circular references.

It raises JSONEncodeError if a tzinfo on a datetime object is unsupported.

JSONEncodeError is a subclass of TypeError. This is for compatibility with the standard library.


To serialize a subclass or arbitrary types, specify default as a callable that returns a supported type. default may be a function, lambda, or callable class instance. To specify that a type was not handled by default, raise an exception such as TypeError.

>>> import orjson, decimal
def default(obj):
    if isinstance(obj, decimal.Decimal):
        return str(obj)
    raise TypeError

>>> orjson.dumps(decimal.Decimal("0.0842389659712649442845"))
JSONEncodeError: Type is not JSON serializable: decimal.Decimal
>>> orjson.dumps(decimal.Decimal("0.0842389659712649442845"), default=default)
>>> orjson.dumps({1, 2}, default=default)
orjson.JSONEncodeError: Type is not JSON serializable: set

The default callable may return an object that itself must be handled by default up to 254 times before an exception is raised.

It is important that default raise an exception if a type cannot be handled. Python otherwise implicitly returns None, which appears to the caller like a legitimate value and is serialized:

>>> import orjson, json, rapidjson
def default(obj):
    if isinstance(obj, decimal.Decimal):
        return str(obj)

>>> orjson.dumps({"set":{1, 2}}, default=default)
>>> json.dumps({"set":{1, 2}}, default=default)
>>> rapidjson.dumps({"set":{1, 2}}, default=default)


To modify how data is serialized, specify option. Each option is an integer constant in orjson. To specify multiple options, mask them together, e.g., option=orjson.OPT_STRICT_INTEGER | orjson.OPT_NAIVE_UTC.


Append \n to the output. This is a convenience and optimization for the pattern of dumps(...) + "\n". bytes objects are immutable and this pattern copies the original contents.

>>> import orjson
>>> orjson.dumps([])
>>> orjson.dumps([], option=orjson.OPT_APPEND_NEWLINE)

Pretty-print output with an indent of two spaces. This is equivalent to indent=2 in the standard library. Pretty printing is slower and the output larger. orjson is the fastest compared library at pretty printing and has much less of a slowdown to pretty print than the standard library does. This option is compatible with all other options.

>>> import orjson
>>> orjson.dumps({"a": "b", "c": {"d": True}, "e": [1, 2]})
>>> orjson.dumps(
    {"a": "b", "c": {"d": True}, "e": [1, 2]},
b'{\n  "a": "b",\n  "c": {\n    "d": true\n  },\n  "e": [\n    1,\n    2\n  ]\n}'

If displayed, the indentation and linebreaks appear like this:

  "a": "b",
  "c": {
    "d": true
  "e": [

This measures serializing the github.json fixture as compact (52KiB) or pretty (64KiB):

Library compact (ms) pretty (ms) vs. orjson
orjson 0.06 0.07 1.0
ujson 0.18 0.19 2.8
rapidjson 0.22
simplejson 0.35 1.49 21.4
json 0.36 1.19 17.2

This measures serializing the citm_catalog.json fixture, more of a worst case due to the amount of nesting and newlines, as compact (489KiB) or pretty (1.1MiB):

Library compact (ms) pretty (ms) vs. orjson
orjson 0.88 1.73 1.0
ujson 3.73 4.52 2.6
rapidjson 3.54
simplejson 11.77 72.06 41.6
json 6.71 55.22 31.9

rapidjson is blank because it does not support pretty printing. This can be reproduced using the pyindent script.


Serialize datetime.datetime objects without a tzinfo as UTC. This has no effect on datetime.datetime objects that have tzinfo set.

>>> import orjson, datetime
>>> orjson.dumps(
        datetime.datetime(1970, 1, 1, 0, 0, 0),
>>> orjson.dumps(
        datetime.datetime(1970, 1, 1, 0, 0, 0),

Serialize dict keys of type other than str. This allows dict keys to be one of str, int, float, bool, None, datetime.datetime,, datetime.time, enum.Enum, and uuid.UUID. For comparison, the standard library serializes str, int, float, bool or None by default. orjson benchmarks as being faster at serializing non-str keys than other libraries. This option is slower for str keys than the default.

>>> import orjson, datetime, uuid
>>> orjson.dumps(
        {uuid.UUID("7202d115-7ff3-4c81-a7c1-2a1f067b1ece"): [1, 2, 3]},
>>> orjson.dumps(
        {datetime.datetime(1970, 1, 1, 0, 0, 0): [1, 2, 3]},
        option=orjson.OPT_NON_STR_KEYS | orjson.OPT_NAIVE_UTC,

These types are generally serialized how they would be as values, e.g., datetime.datetime is still an RFC 3339 string and respects options affecting it. The exception is that int serialization does not respect OPT_STRICT_INTEGER.

This option has the risk of creating duplicate keys. This is because non-str objects may serialize to the same str as an existing key, e.g., {"1": true, 1: false}. The last key to be inserted to the dict will be serialized last and a JSON deserializer will presumably take the last occurrence of a key (in the above, false). The first value will be lost.

This option is compatible with orjson.OPT_SORT_KEYS. If sorting is used, note the sort is unstable and will be unpredictable for duplicate keys.

>>> import orjson, datetime
>>> orjson.dumps(
    {"other": 1,, 1, 5): 2,, 1, 3): 3},
    option=orjson.OPT_NON_STR_KEYS | orjson.OPT_SORT_KEYS

This measures serializing 589KiB of JSON comprising a list of 100 dict in which each dict has both 365 randomly-sorted int keys representing epoch timestamps as well as one str key and the value for each key is a single integer. In "str keys", the keys were converted to str before serialization, and orjson still specifes option=orjson.OPT_NON_STR_KEYS (which is always somewhat slower).

Library str keys (ms) int keys (ms) int keys sorted (ms)
orjson 1.53 2.16 4.29
ujson 3.07 5.65
rapidjson 4.29
simplejson 11.24 14.50 21.86
json 7.17 8.49

ujson is blank for sorting because it segfaults. json is blank because it raises TypeError on attempting to sort before converting all keys to str. rapidjson is blank because it does not support non-str keys. This can be reproduced using the pynonstr script.


Do not serialize the microsecond field on datetime.datetime and datetime.time instances.

>>> import orjson, datetime
>>> orjson.dumps(
        datetime.datetime(1970, 1, 1, 0, 0, 0, 1),
>>> orjson.dumps(
        datetime.datetime(1970, 1, 1, 0, 0, 0, 1),

Passthrough dataclasses.dataclass instances to default. This allows customizing their output but is much slower.

>>> import orjson, dataclasses
class User:
    id: str
    name: str
    password: str

def default(obj):
    if isinstance(obj, User):
        return {"id":, "name":}
    raise TypeError

>>> orjson.dumps(User("3b1", "asd", "zxc"))
>>> orjson.dumps(User("3b1", "asd", "zxc"), option=orjson.OPT_PASSTHROUGH_DATACLASS)
TypeError: Type is not JSON serializable: User
>>> orjson.dumps(
        User("3b1", "asd", "zxc"),

Passthrough datetime.datetime,, and datetime.time instances to default. This allows serializing datetimes to a custom format, e.g., HTTP dates:

>>> import orjson, datetime
def default(obj):
    if isinstance(obj, datetime.datetime):
        return obj.strftime("%a, %d %b %Y %H:%M:%S GMT")
    raise TypeError

>>> orjson.dumps({"created_at": datetime.datetime(1970, 1, 1)})
>>> orjson.dumps({"created_at": datetime.datetime(1970, 1, 1)}, option=orjson.OPT_PASSTHROUGH_DATETIME)
TypeError: Type is not JSON serializable: datetime.datetime
>>> orjson.dumps(
        {"created_at": datetime.datetime(1970, 1, 1)},
b'{"created_at":"Thu, 01 Jan 1970 00:00:00 GMT"}'

This does not affect datetimes in dict keys if using OPT_NON_STR_KEYS.


Passthrough subclasses of builtin types to default.

>>> import orjson
class Secret(str):

def default(obj):
    if isinstance(obj, Secret):
        return "******"
    raise TypeError

>>> orjson.dumps(Secret("zxc"))
>>> orjson.dumps(Secret("zxc"), option=orjson.OPT_PASSTHROUGH_SUBCLASS)
TypeError: Type is not JSON serializable: Secret
>>> orjson.dumps(Secret("zxc"), option=orjson.OPT_PASSTHROUGH_SUBCLASS, default=default)

This does not affect serializing subclasses as dict keys if using OPT_NON_STR_KEYS.


This is deprecated and has no effect in version 3. In version 2 this was required to serialize dataclasses.dataclass instances. For more, see dataclass.


Serialize numpy.ndarray instances. For more, see numpy.


This is deprecated and has no effect in version 3. In version 2 this was required to serialize uuid.UUID instances. For more, see UUID.


Serialize dict keys in sorted order. The default is to serialize in an unspecified order. This is equivalent to sort_keys=True in the standard library.

This can be used to ensure the order is deterministic for hashing or tests. It has a substantial performance penalty and is not recommended in general.

>>> import orjson
>>> orjson.dumps({"b": 1, "c": 2, "a": 3})
>>> orjson.dumps({"b": 1, "c": 2, "a": 3}, option=orjson.OPT_SORT_KEYS)

This measures serializing the twitter.json fixture unsorted and sorted:

Library unsorted (ms) sorted (ms) vs. orjson
orjson 0.5 0.92 1
ujson 1.61 2.48 2.7
rapidjson 2.17 2.89 3.2
simplejson 3.56 5.13 5.6
json 3.59 4.59 5

The benchmark can be reproduced using the pysort script.

The sorting is not collation/locale-aware:

>>> import orjson
>>> orjson.dumps({"a": 1, "ä": 2, "A": 3}, option=orjson.OPT_SORT_KEYS)

This is the same sorting behavior as the standard library, rapidjson, simplejson, and ujson.

dataclass also serialize as maps but this has no effect on them.


Enforce 53-bit limit on integers. The limit is otherwise 64 bits, the same as the Python standard library. For more, see int.


Serialize a UTC timezone on datetime.datetime instances as Z instead of +00:00.

>>> import orjson, datetime
>>> orjson.dumps(
        datetime.datetime(1970, 1, 1, 0, 0, 0, tzinfo=datetime.timezone.utc),
>>> orjson.dumps(
        datetime.datetime(1970, 1, 1, 0, 0, 0, tzinfo=datetime.timezone.utc),


def loads(__obj: Union[bytes, bytearray, str]) -> Any: ...

loads() deserializes JSON to Python objects. It deserializes to dict, list, int, float, str, bool, and None objects.

bytes, bytearray, and str input are accepted. If the input exists as bytes (was read directly from a source), it is recommended to pass bytes. This has lower memory usage and lower latency.

The input must be valid UTF-8.

orjson maintains a cache of map keys for the duration of the process. This causes a net reduction in memory usage by avoiding duplicate strings. The keys must be at most 64 bytes to be cached and 512 entries are stored.

The global interpreter lock (GIL) is held for the duration of the call.

It raises JSONDecodeError if given an invalid type or invalid JSON. This includes if the input contains NaN, Infinity, or -Infinity, which the standard library allows, but is not valid JSON.

JSONDecodeError is a subclass of json.JSONDecodeError and ValueError. This is for compatibility with the standard library.



orjson serializes instances of dataclasses.dataclass natively. It serializes instances 40-50x as fast as other libraries and avoids a severe slowdown seen in other libraries compared to serializing dict.

It is supported to pass all variants of dataclasses, including dataclasses using __slots__, frozen dataclasses, those with optional or default attributes, and subclasses. There is a performance benefit to not using __slots__.

Library dict (ms) dataclass (ms) vs. orjson
orjson 1.40 1.60 1
rapidjson 3.64 68.48 42
simplejson 14.21 92.18 57
json 13.28 94.90 59

This measures serializing 555KiB of JSON, orjson natively and other libraries using default to serialize the output of dataclasses.asdict(). This can be reproduced using the pydataclass script.

Dataclasses are serialized as maps, with every attribute serialized and in the order given on class definition:

>>> import dataclasses, orjson, typing

class Member:
    id: int
    active: bool = dataclasses.field(default=False)

class Object:
    id: int
    name: str
    members: typing.List[Member]

>>> orjson.dumps(Object(1, "a", [Member(1, True), Member(2)]))

Users may wish to control how dataclass instances are serialized, e.g., to not serialize an attribute or to change the name of an attribute when serialized. orjson may implement support using the metadata mapping on field attributes, e.g., field(metadata={"json_serialize": False}), if use cases are clear.


orjson serializes datetime.datetime objects to RFC 3339 format, e.g., "1970-01-01T00:00:00+00:00". This is a subset of ISO 8601 and compatible with isoformat() in the standard library.

>>> import orjson, datetime, zoneinfo
>>> orjson.dumps(
    datetime.datetime(2018, 12, 1, 2, 3, 4, 9, tzinfo=zoneinfo.ZoneInfo('Australia/Adelaide'))
>>> orjson.dumps(
>>> orjson.dumps(

datetime.datetime supports instances with a tzinfo that is None, datetime.timezone.utc, a timezone instance from the python3.9+ zoneinfo module, or a timezone instance from the third-party pendulum, pytz, or dateutil/arrow libraries.

datetime.time objects must not have a tzinfo.

>>> import orjson, datetime
>>> orjson.dumps(datetime.time(12, 0, 15, 290))
b'"12:00:15.000290"' objects will always serialize.

>>> import orjson, datetime
>>> orjson.dumps(, 1, 2))

Errors with tzinfo result in JSONEncodeError being raised.

It is faster to have orjson serialize datetime objects than to do so before calling dumps(). If using an unsupported type such as pendulum.datetime, use default.

To disable serialization of datetime objects specify the option orjson.OPT_PASSTHROUGH_DATETIME.


orjson serializes enums natively. Options apply to their values.

>>> import enum, datetime, orjson
class DatetimeEnum(enum.Enum):
    EPOCH = datetime.datetime(1970, 1, 1, 0, 0, 0)
>>> orjson.dumps(DatetimeEnum.EPOCH)
>>> orjson.dumps(DatetimeEnum.EPOCH, option=orjson.OPT_NAIVE_UTC)

Enums with members that are not supported types can be serialized using default:

>>> import enum, orjson
class Custom:
    def __init__(self, val):
        self.val = val

def default(obj):
    if isinstance(obj, Custom):
        return obj.val
    raise TypeError

class CustomEnum(enum.Enum):
    ONE = Custom(1)

>>> orjson.dumps(CustomEnum.ONE, default=default)


orjson serializes and deserializes double precision floats with no loss of precision and consistent rounding. The same behavior is observed in rapidjson, simplejson, and json. ujson 1.35 was inaccurate in both serialization and deserialization, i.e., it modifies the data, and the recent 2.0 release is accurate.

orjson.dumps() serializes Nan, Infinity, and -Infinity, which are not compliant JSON, as null:

>>> import orjson, ujson, rapidjson, json
>>> orjson.dumps([float("NaN"), float("Infinity"), float("-Infinity")])
>>> ujson.dumps([float("NaN"), float("Infinity"), float("-Infinity")])
OverflowError: Invalid Inf value when encoding double
>>> rapidjson.dumps([float("NaN"), float("Infinity"), float("-Infinity")])
>>> json.dumps([float("NaN"), float("Infinity"), float("-Infinity")])
'[NaN, Infinity, -Infinity]'


orjson serializes and deserializes 64-bit integers by default. The range supported is a signed 64-bit integer's minimum (-9223372036854775807) to an unsigned 64-bit integer's maximum (18446744073709551615). This is widely compatible, but there are implementations that only support 53-bits for integers, e.g., web browsers. For those implementations, dumps() can be configured to raise a JSONEncodeError on values exceeding the 53-bit range.

>>> import orjson
>>> orjson.dumps(9007199254740992)
>>> orjson.dumps(9007199254740992, option=orjson.OPT_STRICT_INTEGER)
JSONEncodeError: Integer exceeds 53-bit range
>>> orjson.dumps(-9007199254740992, option=orjson.OPT_STRICT_INTEGER)
JSONEncodeError: Integer exceeds 53-bit range


orjson natively serializes numpy.ndarray and individual numpy.float64, numpy.float32, numpy.int64, numpy.int32, numpy.int8, numpy.uint64, numpy.uint32, and numpy.uint8 instances. Arrays may have a dtype of numpy.bool, numpy.float32, numpy.float64, numpy.int32, numpy.int64, numpy.uint32, numpy.uint64, numpy.uintp, or numpy.intp. orjson is faster than all compared libraries at serializing numpy instances. Serializing numpy data requires specifying option=orjson.OPT_SERIALIZE_NUMPY.

>>> import orjson, numpy
>>> orjson.dumps(
        numpy.array([[1, 2, 3], [4, 5, 6]]),

The array must be a contiguous C array (C_CONTIGUOUS) and one of the supported datatypes.

If an array is not a contiguous C array or contains an supported datatype, orjson falls through to default. In default, obj.tolist() can be specified. If an array is malformed, which is not expected, orjson.JSONEncodeError is raised.

This measures serializing 92MiB of JSON from an numpy.ndarray with dimensions of (50000, 100) and numpy.float64 values:

Library Latency (ms) RSS diff (MiB) vs. orjson
orjson 194 99 1.0
rapidjson 3,048 309 15.7
simplejson 3,023 297 15.6
json 3,133 297 16.1

This measures serializing 100MiB of JSON from an numpy.ndarray with dimensions of (100000, 100) and numpy.int32 values:

Library Latency (ms) RSS diff (MiB) vs. orjson
orjson 178 115 1.0
rapidjson 1,512 551 8.5
simplejson 1,606 504 9.0
json 1,506 503 8.4

This measures serializing 105MiB of JSON from an numpy.ndarray with dimensions of (100000, 200) and numpy.bool values:

Library Latency (ms) RSS diff (MiB) vs. orjson
orjson 157 120 1.0
rapidjson 710 327 4.5
simplejson 931 398 5.9
json 996 400 6.3

In these benchmarks, orjson serializes natively, ujson is blank because it does not support a default parameter, and the other libraries serialize ndarray.tolist() via default. The RSS column measures peak memory usage during serialization. This can be reproduced using the pynumpy script.

orjson does not have an installation or compilation dependency on numpy. The implementation is independent, reading numpy.ndarray using PyArrayInterface.


orjson is strict about UTF-8 conformance. This is stricter than the standard library's json module, which will serialize and deserialize UTF-16 surrogates, e.g., "\ud800", that are invalid UTF-8.

If orjson.dumps() is given a str that does not contain valid UTF-8, orjson.JSONEncodeError is raised. If loads() receives invalid UTF-8, orjson.JSONDecodeError is raised.

orjson and rapidjson are the only compared JSON libraries to consistently error on bad input.

>>> import orjson, ujson, rapidjson, json
>>> orjson.dumps('\ud800')
JSONEncodeError: str is not valid UTF-8: surrogates not allowed
>>> ujson.dumps('\ud800')
UnicodeEncodeError: 'utf-8' codec ...
>>> rapidjson.dumps('\ud800')
UnicodeEncodeError: 'utf-8' codec ...
>>> json.dumps('\ud800')
>>> orjson.loads('"\\ud800"')
JSONDecodeError: unexpected end of hex escape at line 1 column 8: line 1 column 1 (char 0)
>>> ujson.loads('"\\ud800"')
>>> rapidjson.loads('"\\ud800"')
ValueError: Parse error at offset 1: The surrogate pair in string is invalid.
>>> json.loads('"\\ud800"')

To make a best effort at deserializing bad input, first decode bytes using the replace or lossy argument for errors:

>>> import orjson
>>> orjson.loads(b'"\xed\xa0\x80"')
JSONDecodeError: str is not valid UTF-8: surrogates not allowed
>>> orjson.loads(b'"\xed\xa0\x80"'.decode("utf-8", "replace"))


orjson serializes uuid.UUID instances to RFC 4122 format, e.g., "f81d4fae-7dec-11d0-a765-00a0c91e6bf6".

>>> import orjson, uuid
>>> orjson.dumps(uuid.UUID('f81d4fae-7dec-11d0-a765-00a0c91e6bf6'))
>>> orjson.dumps(uuid.uuid5(uuid.NAMESPACE_DNS, ""))


The library has comprehensive tests. There are tests against fixtures in the JSONTestSuite and nativejson-benchmark repositories. It is tested to not crash against the Big List of Naughty Strings. It is tested to not leak memory. It is tested to not crash against and not accept invalid UTF-8. There are integration tests exercising the library's use in web servers (gunicorn using multiprocess/forked workers) and when multithreaded. It also uses some tests from the ultrajson library.

orjson is the most correct of the compared libraries. This graph shows how each library handles a combined 342 JSON fixtures from the JSONTestSuite and nativejson-benchmark tests:

Library Invalid JSON documents not rejected Valid JSON documents not deserialized
orjson 0 0
ujson 38 0
rapidjson 6 0
simplejson 13 0
json 17 0

This shows that all libraries deserialize valid JSON but only orjson correctly rejects the given invalid JSON fixtures. Errors are largely due to accepting invalid strings and numbers.

The graph above can be reproduced using the pycorrectness script.


Serialization and deserialization performance of orjson is better than ultrajson, rapidjson, simplejson, or json. The benchmarks are done on fixtures of real data:

  • twitter.json, 631.5KiB, results of a search on Twitter for "一", containing CJK strings, dictionaries of strings and arrays of dictionaries, indented.

  • github.json, 55.8KiB, a GitHub activity feed, containing dictionaries of strings and arrays of dictionaries, not indented.

  • citm_catalog.json, 1.7MiB, concert data, containing nested dictionaries of strings and arrays of integers, indented.

  • canada.json, 2.2MiB, coordinates of the Canadian border in GeoJSON format, containing floats and arrays, indented.


alt text alt text alt text alt text alt text alt text alt text alt text

twitter.json serialization

Library Median latency (milliseconds) Operations per second Relative (latency)
orjson 0.59 1698.8 1
ujson 2.14 464.3 3.64
rapidjson 2.39 418.5 4.06
simplejson 3.15 316.9 5.36
json 3.56 281.2 6.06

twitter.json deserialization

Library Median latency (milliseconds) Operations per second Relative (latency)
orjson 2.28 439.3 1
ujson 2.89 345.9 1.27
rapidjson 3.85 259.6 1.69
simplejson 3.66 272.1 1.61
json 4.05 246.7 1.78

github.json serialization

Library Median latency (milliseconds) Operations per second Relative (latency)
orjson 0.07 15265.2 1
ujson 0.22 4556.7 3.35
rapidjson 0.26 3808.9 4.02
simplejson 0.37 2690.4 5.68
json 0.35 2847.8 5.36

github.json deserialization

Library Median latency (milliseconds) Operations per second Relative (latency)
orjson 0.18 5610.1 1
ujson 0.28 3540.7 1.58
rapidjson 0.33 3031.5 1.85
simplejson 0.29 3385.6 1.65
json 0.29 3402.1 1.65

citm_catalog.json serialization

Library Median latency (milliseconds) Operations per second Relative (latency)
orjson 0.99 1008.5 1
ujson 3.69 270.7 3.72
rapidjson 3.55 281.4 3.58
simplejson 11.76 85.1 11.85
json 6.89 145.1 6.95

citm_catalog.json deserialization

Library Median latency (milliseconds) Operations per second Relative (latency)
orjson 4.53 220.5 1
ujson 5.67 176.5 1.25
rapidjson 7.51 133.3 1.66
simplejson 7.54 132.7 1.66
json 7.8 128.2 1.72

canada.json serialization

Library Median latency (milliseconds) Operations per second Relative (latency)
orjson 4.72 198.9 1
ujson 17.76 56.3 3.77
rapidjson 61.83 16.2 13.11
simplejson 80.6 12.4 17.09
json 52.38 18.8 11.11

canada.json deserialization

Library Median latency (milliseconds) Operations per second Relative (latency)
orjson 10.28 97.4 1
ujson 16.49 60.5 1.6
rapidjson 37.92 26.4 3.69
simplejson 37.7 26.5 3.67
json 37.87 27.6 3.68


orjson's memory usage when deserializing is similar to or lower than the standard library and other third-party libraries.

This measures, in the first column, RSS after importing a library and reading the fixture, and in the second column, increases in RSS after repeatedly calling loads() on the fixture.


Library import, read() RSS (MiB) loads() increase in RSS (MiB)
orjson 13.5 2.5
ujson 14 4.1
rapidjson 14.7 6.5
simplejson 13.2 2.5
json 12.9 2.3


Library import, read() RSS (MiB) loads() increase in RSS (MiB)
orjson 13.1 0.3
ujson 13.5 0.3
rapidjson 14 0.7
simplejson 12.6 0.3
json 12.3 0.1


Library import, read() RSS (MiB) loads() increase in RSS (MiB)
orjson 14.6 7.9
ujson 15.1 11.1
rapidjson 15.8 36
simplejson 14.3 27.4
json 14 27.2


Library import, read() RSS (MiB) loads() increase in RSS (MiB)
orjson 17.1 15.7
ujson 17.6 17.4
rapidjson 18.3 17.9
simplejson 16.9 19.6
json 16.5 19.4


The above was measured using Python 3.8.3 on Linux (x86_64) with orjson 3.3.0, ujson 3.0.0, python-rapidson 0.9.1, and simplejson 3.17.2.

The latency results can be reproduced using the pybench and graph scripts. The memory results can be reproduced using the pymem script.


Why can't I install it from PyPI?

Probably pip needs to be upgraded. pip added support for manylinux2014 in 2019.

Will it deserialize to dataclasses, UUIDs, decimals, etc or support object_hook?

No. This requires a schema specifying what types are expected and how to handle errors etc. This is addressed by data validation libraries a level above this.

Will it serialize to str?

No. bytes is the correct type for a serialized blob.

Will it support PyPy?

If someone implements it well.


To package orjson requires Rust on the nightly channel and the maturin build tool. maturin can be installed from PyPI or packaged as well. This is the simplest and recommended way of installing from source, assuming rustup is available from a package manager:

rustup default nightly
pip wheel --no-binary=orjson orjson

This is an example of building a wheel using the repository as source, rustup installed from upstream, and a pinned version of Rust:

pip install maturin
curl -sSf | sh -s -- --default-toolchain nightly-2021-01-31 --profile minimal -y
maturin build --no-sdist --release --strip --manylinux off
ls -1 target/wheels

Problems with the Rust nightly channel may require pinning a version. nightly-2021-01-31 is known to be ok.

orjson is tested for amd64 and aarch64 on Linux, macOS, and Windows. It may not work on 32-bit targets. It has recommended RUSTFLAGS specified in .cargo/config so it is recommended to either not set RUSTFLAGS or include these options.

There are no runtime dependencies other than libc.

orjson's tests are included in the source distribution on PyPI. It is necessarily to install dependencies from PyPI specified in test/requirements.txt. These require a C compiler. The tests do not make network requests.

The tests should be run as part of the build. It can be run like this:

pip install -r test/requirements.txt
pytest -q test


orjson was written by ijl <[email protected]>, copyright 2018 - 2021, licensed under both the Apache 2 and MIT licenses.

  • Add option to serialize enums as their names

    Add option to serialize enums as their names

    This implements #207
    An option to serialize enums as their names instead of values.

    class Foo(enum.Enum):
        ONE = 1
    print(orjson.dumps(Foo.ONE, option=orjson.OPT_ENUM_NAME))

    Which will print b'"ONE"'

    Right now the AutoEnum test fails because enums created by are not handled by the enum serialize code.
    I am not sure what the best solution is, since these enums also do not have a name attribute.
    Maybe this option/test should just ignore enums?

    opened by ydylla 20
  • Relax strict enforcing of keys type to string

    Relax strict enforcing of keys type to string

    I would like to use orjson for structlog, but as it turns out, people are dumping arbitrary data for debugging purposes, where you cannot ensure dict key types are always strings.

    So I ask, if there is even a technical reason to enforce string keys and if not, if it is possible to optionally relax this constraint...?

    opened by diefans 20
  • Feature: serialize UUIDs natively

    Feature: serialize UUIDs natively

    Closes #41

    This is an initial, clumsy, incorrect implementation -- I'm not even sure how to run the tests at this point, and I know that converting the 128-bit value in the UUID can't be as easy as calling one C function.

    But I'd really appreciate some :eyes: and review while I figure out the rest (tips on what to do also appreciated ;-))

    opened by necaris 15
  • Cannot pip install orjson (python 3.6).

    Cannot pip install orjson (python 3.6).

    sudo pip3 install --upgrade orjson:
    Complete output from command python egg_info:
        Traceback (most recent call last):
          File "<string>", line 1, in <module>
          File "/usr/lib/python3.6/", line 452, in open
            buffer = _builtin_open(filename, 'rb')
        FileNotFoundError: [Errno 2] No such file or directory: '/tmp/pip-build-3dwbn4b3/orjson/'

    Not sure why it can't find setup when it was just downloaded. I've verified that it exists in the source.

    opened by loomcode 13
  • Apple Silicon Binaries

    Apple Silicon Binaries

    Hi all,

    Thanks for this great library! I found that it is a little difficult (but not too difficult) to install on Apple Silicon as it requires installing rust nightly to get maturin installed.

    Would it be possible to release binaries for this platform?


    help wanted 
    opened by william-silversmith 12
  • add support for Pydantic's `BaseModel`

    add support for Pydantic's `BaseModel`

    We have some use cases of calling model.dict() then dumping to json. Pydantic's dict is very slow and we rather just call orjson.dumps(model). It uses the same serialization of dataclass, as it seemed like same usecase.

    opened by aviramha 11
  • 3.6.6 silent crash on windows (packed by pyinstaller)

    3.6.6 silent crash on windows (packed by pyinstaller)

    we're using orjson within a pretty-complex application (along with pyside6 and other huge libraries). the final product is being packaged for windows using pyinstaller.

    lately, our app fails to start. there's some cryptic message about application crash in windows event log. after investigating, it seems the crash is caused by orjson-3.6.6 - when built with 3.6.5 (everything else being the same), it runs correctly.

    i don't have much time to investigate further right now. hopefully you'll have the idea what might be wrong... ...if not, i'll try to provide more info later...

    opened by rpodgorny 10
  • Support for Subclasses of builtins, Enums

    Support for Subclasses of builtins, Enums

    Hello -

    I'm very interested in making use of this library for its native support of dataclasses, however I have a few minor blockers:

    1. I make use of enum.Enum subclasses quite extensively,
    2. I also make use of subclasses of builtins for data-validation.

    Both of these cases are natively handled by the builtin json library and other third-party software (rapidjson, ujson, etc). I think perhaps supporting subclasses of builtins would solve both of my two blockers.

    I also understand that I could implement a default callable for these cases. However, in a code-base using almost-exclusively subtypes in this manner, the potential performance gains are lost by constantly bailing out to the default.

    Unfortunately, I'm not familiar enough with Rust or its Python interface to suggest an implementation, but I can provide a test case:

    import enum
    import dataclasses
    import orjson
    class ShortStr(str):
        def __new__(cls, *args, **kwargs):
             v = str.__new__(cls, *args, **kwargs)
             assert len(v) < 10, "Values must be less than 10 chars!"
             return v
    class Choice(int, enum.Enum):
        YES = 1
        NO = 0
        MAYBE = -1
    class UncertaintyPrinciple:
        subject: ShortStr
        decision: Choice
        def json(self) -> str:
            return orjson.dumps(self, option=orjson.OPT_SERIALIZE_DATACLASS)
    opened by seandstewart 9
  • Enable compliant manylinux1 builds on Azure

    Enable compliant manylinux1 builds on Azure

    This PR (hopefully) enables manylinux1 builds by using the konstin2/maturin:master Docker image as outlined in

    I've completely removed Python 3.9 for now, as it is not yet included in the manylinux Docker image. The simplest solution would be to fallback to the previous template for Python 3.9.

    opened by pskopnik 9
  • Alias dataclass field name

    Alias dataclass field name

    When parsing JSON, sometimes the fields need/have horrible names so I think it'd be neat to add an option to specify the JSON field name for encoding/decoding.


    class A:
        b: int = field(metadata={"name": 'anothername'})
    # will print
    b'{"anothername": 1}'
    opened by aviramha 9
  • ImportError on CentOS 7

    ImportError on CentOS 7

    Installed with:

    pip3 install --user --upgrade orjson

    Fails on import:

    import orjson
    ImportError: /lib64/ version `GLIBC_2.18' not found (required by /home/davidgaleano/.local/lib/python3.6/site-packages/

    The python version is 3.6.3, and the CentOS version is 7.4.1708.

    opened by davidgaleano 9
  • Result of dumps is corrupt in 3.8.4 (windows only)

    Result of dumps is corrupt in 3.8.4 (windows only)

    After updating to 3.8.4, I have seen some errors. The dumps results are giving "JSONDecodeError" and not decoding correctly as JSON. Revert to 3.8.3, I don't have this problem.

    I have tried it in Windows and Mac environments, but it only occurs in Windows. Therefore, it is possible that this is a problem specific to me, but I am reporting it just in case.

    I noticed that some of the data that corrupts the output appears to be random, but has a high probability of being corrupted. I created a simple program to test a file (errors.json) that collected them.

    import orjson
    idx = 0
    count = 0
    with open('errors.json', 'rt', encoding='utf-8') as f:
        for line in f:
            except orjson.JSONDecodeError:
                print(f'line.#{idx} JSONDecodeError')
                count += 1
            idx += 1
    print(f'{count}/{idx} error happen.')
    PS R:\orjson_tests> poetry run python .\
    line.#4 JSONDecodeError
    line.#8 JSONDecodeError
    line.#16 JSONDecodeError
    line.#17 JSONDecodeError
    line.#19 JSONDecodeError
    line.#21 JSONDecodeError
    line.#23 JSONDecodeError
    line.#24 JSONDecodeError
    line.#25 JSONDecodeError
    line.#28 JSONDecodeError
    line.#29 JSONDecodeError
    line.#30 JSONDecodeError
    line.#31 JSONDecodeError
    line.#32 JSONDecodeError
    line.#33 JSONDecodeError
    line.#34 JSONDecodeError
    line.#36 JSONDecodeError
    line.#37 JSONDecodeError
    line.#38 JSONDecodeError
    line.#39 JSONDecodeError
    line.#41 JSONDecodeError
    line.#43 JSONDecodeError
    line.#46 JSONDecodeError
    line.#47 JSONDecodeError
    line.#49 JSONDecodeError
    line.#54 JSONDecodeError
    line.#55 JSONDecodeError
    line.#59 JSONDecodeError
    line.#61 JSONDecodeError
    29/69 error happen.

    Python version is 3.11.1 / Windows11

    opened by ryuuji 3
  • 3.8.4(Jan 4, 2023)

  • 3.8.3(Dec 2, 2022)

  • 3.8.2(Nov 20, 2022)


    • Fix tests on 32-bit for numpy.intp and numpy.uintp.


    • Build now depends on rustc 1.60 or later.
    • Support building with maturin 0.13 or 0.14.
    Source code(tar.gz)
    Source code(zip)
  • 3.8.1(Oct 25, 2022)

  • 3.8.0(Aug 27, 2022)

  • 3.7.12(Aug 14, 2022)

  • 3.7.11(Jul 31, 2022)

  • 3.7.10(Jul 30, 2022)

  • 3.7.9(Jul 29, 2022)


    • Improve performance of serializing.
    • Improve performance of serializing pretty-printed (orjson.OPT_INDENT_2) to be much nearer to compact.
    • Improve performance of deserializing str input.
    • orjson now requires Rust 1.57 instead of 1.54 to build.
    Source code(tar.gz)
    Source code(zip)
  • 3.7.8(Jul 19, 2022)

  • 3.7.7(Jul 6, 2022)

  • 3.7.6(Jul 3, 2022)

  • 3.7.5(Jun 28, 2022)

  • 3.7.4(Jun 28, 2022)

  • 3.7.3(Jun 23, 2022)

  • 3.7.2(Jun 7, 2022)

  • 3.7.1(Jun 3, 2022)


    • Type stubs for orjson.JSONDecodeError now inherit from json.JSONDecodeError instead of ValueError
    • Null-terminate the internal buffer of orjson.dumps() output.
    Source code(tar.gz)
    Source code(zip)
  • 3.7.0(Jun 3, 2022)


    • Improve deserialization performance significantly through the use of a new backend. PyPI wheels for manylinux_2_28 and macOS have it enabled. Packagers are advised to see the README.
    Source code(tar.gz)
    Source code(zip)
  • 3.6.9(Jun 1, 2022)

  • 3.6.8(Apr 15, 2022)

  • 3.6.7(Feb 14, 2022)


    • Improve performance of deserializing almost-empty documents.
    • Publish arm7l manylinux_2_17 wheels to PyPI.
    • Publish amd64 musllinux_1_1 wheels to PyPI.


    • Fix build requiring python on PATH.
    Source code(tar.gz)
    Source code(zip)
  • 3.6.6(Jan 21, 2022)


    • Improve performance of serializing datetime.datetime using tzinfo that are zoneinfo.ZoneInfo.


    • Fix invalid indexing in line and column number reporting in JSONDecodeError.
    • Fix orjson.OPT_STRICT_INTEGER not raising an error on values exceeding a 64-bit integer maximum.
    Source code(tar.gz)
    Source code(zip)
  • 3.6.5(Dec 5, 2021)

  • 3.6.4(Oct 1, 2021)


    • Fix serialization of dataclass inheriting from abc.ABC and using __slots__.
    • Decrement refcount for numpy PyArrayInterface.
    • Fix build on recent versions of Rust nightly.
    Source code(tar.gz)
    Source code(zip)
  • 3.6.3(Aug 20, 2021)

  • 3.6.2(Aug 17, 2021)


    • orjson now compiles on Rust stable 1.54.0 or above. Use of some SIMD usage is now disabled by default and packagers are advised to add --cargo-extra-args="--features=unstable-simd" to the maturin build command if they continue to use nightly.
    • orjson built with --features=unstable-simd adds UTF-8 validation implementations that use AVX2 or SSE4.2.
    • Drop support for Python 3.6.
    Source code(tar.gz)
    Source code(zip)
  • 3.6.1(Aug 4, 2021)


    • orjson now includes a pyi type stubs file.
    • Publish manylinux_2_24 wheels instead of manylinux2014.


    • Fix compilation on latest Rust nightly.
    Source code(tar.gz)
    Source code(zip)
  • 3.6.0(Jul 8, 2021)

  • 3.5.4(Jun 30, 2021)


    • Fix memory leak serializing datetime.datetime with tzinfo.
    • Fix wrong error message when serializing an unsupported numpy type without default specified.


    • Publish python3.10 and python3.9 manylinux_2_24 wheels.
    Source code(tar.gz)
    Source code(zip)
  • 3.5.3(Jun 1, 2021)

Ultra fast JSON decoder and encoder written in C with Python bindings

UltraJSON UltraJSON is an ultra fast JSON encoder and decoder written in pure C with bindings for Python 3.6+. Install with pip: $ python -m pip insta

null 3.9k Jan 2, 2023
simplejson is a simple, fast, extensible JSON encoder/decoder for Python

simplejson simplejson is a simple, fast, complete, correct and extensible JSON <> encoder and decoder for Python 3.3+ with legacy suppo

null 1.5k Dec 31, 2022
Crappy tool to convert .scw files to .json and and vice versa.

SCW-JSON-TOOL Crappy tool to convert .scw files to .json and vice versa. How to use Run file with two arguments: python <scw2json or j

Fred31 5 May 14, 2021
🦉 Modern high-performance serialization utilities for Python (JSON, MessagePack, Pickle)

srsly: Modern high-performance serialization utilities for Python This package bundles some of the best Python serialization libraries into one standa

Explosion 329 Dec 28, 2022
A lightweight library for converting complex objects to and from simple Python datatypes.

marshmallow: simplified object serialization marshmallow is an ORM/ODM/framework-agnostic library for converting complex datatypes, such as objects, t

marshmallow-code 6.4k Jan 2, 2023
Generic ASN.1 library for Python

ASN.1 library for Python This is a free and open source implementation of ASN.1 types and codecs as a Python package. It has been first written to sup

Ilya Etingof 223 Dec 11, 2022
FlatBuffers: Memory Efficient Serialization Library

FlatBuffers FlatBuffers is a cross platform serialization library architected for maximum memory efficiency. It allows you to directly access serializ

Google 19.6k Jan 1, 2023
MessagePack serializer implementation for Python[Python]

MessagePack for Python What's this MessagePack is an efficient binary serialization format. It lets you exchange data among multiple languages like JS

MessagePack 1.7k Dec 29, 2022
Extended pickling support for Python objects

cloudpickle cloudpickle makes it possible to serialize Python constructs not supported by the default pickle module from the Python standard library.

null 1.3k Jan 5, 2023
serialize all of python

dill serialize all of python About Dill dill extends python's pickle module for serializing and de-serializing python objects to the majority of the b

The UQ Foundation 1.8k Jan 7, 2023
Python wrapper around rapidjson

python-rapidjson Python wrapper around RapidJSON Authors: Ken Robbins <[email protected]> Lele Gaifax <[email protected]> License: MIT License Sta

null 469 Jan 4, 2023
Python bindings for the simdjson project.

pysimdjson Python bindings for the simdjson project, a SIMD-accelerated JSON parser. If SIMD instructions are unavailable a fallback parser is used, m

Tyler Kennedy 562 Jan 8, 2023
Simple dataclasses configuration management for Python with hocon/json/yaml/properties/env-vars/dict support.

Simple dataclasses configuration management for Python with hocon/json/yaml/properties/env-vars/dict support, based on awesome and lightweight pyhocon parsing library.

Teo Stocco 62 Dec 23, 2022
Define your JSON schema as Python dataclasses

Define your JSON schema as Python dataclasses

null 62 Sep 20, 2022
Python datetimes made easy

Pendulum Python datetimes made easy. Supports Python 2.7 and 3.4+. >>> import pendulum >>> now_in_paris ='Europe/Paris') >>> now_in_par

Sébastien Eustace 5.3k Jan 6, 2023
Datetimes for Humans™

Maya: Datetimes for Humans™ Datetimes are very frustrating to work with in Python, especially when dealing with different locales on different systems

Timo Furrer 3.4k Dec 28, 2022
Datetimes for Humans™

Maya: Datetimes for Humans™ Datetimes are very frustrating to work with in Python, especially when dealing with different locales on different systems

Timo Furrer 3.4k Dec 28, 2022
Typical: Fast, simple, & correct data-validation using Python 3 typing.

typical: Python's Typing Toolkit Introduction Typical is a library devoted to runtime analysis, inference, validation, and enforcement of Python types

Sean 170 Dec 26, 2022
Typical: Fast, simple, & correct data-validation using Python 3 typing.

typical: Python's Typing Toolkit Introduction Typical is a library devoted to runtime analysis, inference, validation, and enforcement of Python types

Sean 171 Jan 2, 2023