List of OpenAI Whisper Checkpoints Variants

Official OpenAI Whisper has 12 unique checkpoint files, but the OpenAI Python package exposes 14 local model names because large and turbo are aliases for large-v3 and large-v3-turbo.

The real confusion comes from aliases, runtimes, and converted formats.

faster-whisper, CTranslate2, whisper.cpp, ggml, q8, q5, ONNX, Core ML, etc. are usually not new Whisper models. They are ways to run or compress the same models.

Official OpenAI Whisper checkpoints

#	Model	Type	Parameters	Note
1	tiny.en	English-only	39M	fastest English model
2	tiny	multilingual	39M	fastest multilingual
3	base.en	English-only	74M	better than tiny
4	base	multilingual	74M	small multilingual baseline
5	small.en	English-only	244M	good lightweight English
6	small	multilingual	244M	good lightweight multilingual
7	medium.en	English-only	769M	strong English model
8	medium	multilingual	769M	strong multilingual model
9	large-v1	multilingual	1550M	original large
10	large-v2	multilingual	1550M	improved large
11	large-v3	multilingual	1550M	best full Whisper model
12	large-v3-turbo	multilingual	~809M	faster optimized large-v3

OpenAI’s README lists the six size families as tiny, base, small, medium, large, and turbo, with .en versions only for the first four families. It also states that turbo is optimized from large-v3 for faster transcription.

The OpenAI Python package accepts 14 names, because two names are aliases:

tiny.en
tiny
base.en
base
small.en
small
medium.en
medium
large-v1
large-v2
large-v3
large
large-v3-turbo
turbo

large = large-v3
turbo = large-v3-turbo

Use .en only when the audio is English. The .en models exist only for tiny, base, small, and medium. There is no large.en, large-v3.en, or turbo.en. OpenAI notes that .en models tend to do better for English, especially at tiny.en and base.en.

Use large-v3 when you want maximum local Whisper accuracy, but test by language/domain.

Use large-v3-turbo or turbo when you want much faster transcription with small accuracy loss.

Do not use turbo for translation.

OpenAI’s official openai-whisper package/CLI explicitly says turbo is not trained for translation and returns the original language even with --task translate.

Use medium, large, large-v2, or large-v3 for translation.

Name you see	What it is	Is it a new Whisper model?
large-v3	official OpenAI Whisper checkpoint	Yes
large-v3-turbo	official OpenAI optimized checkpoint	Yes
turbo	alias for large-v3-turbo	No, alias
large	alias for large-v3 in current OpenAI package	No, alias
whisper-1	OpenAI hosted API model, powered by open-source Whisper V2	API model ID, not a local checkpoint
gpt-4o-transcribe	newer OpenAI transcription model	Not Whisper
gpt-4o-mini-transcribe	newer cheaper transcription model	Not Whisper
faster-whisper	Python implementation using CTranslate2	No
CTranslate2 / ct2	optimized inference format/runtime	No
whisper.cpp	C/C++ implementation using ggml files	No
ggml-*	whisper.cpp model file format	No
q5_0, q5_1, q8_0, int8	quantized/compressed weights	No
ONNX	exported runtime format	No
Core ML	Apple runtime format	No
MLX	Apple Silicon ML runtime format	No
distil-whisper	third-party distilled Whisper-family models	Yes, but not OpenAI official

faster-whisper is not a separate model family. It is a faster implementation of Whisper using CTranslate2, and it can auto-download converted CTranslate2 versions of models. The project describes itself as a reimplementation of OpenAI Whisper using CTranslate2.

Current faster-whisper accepted names include:

tiny.en
tiny
base.en
base
small.en
small
medium.en
medium
large-v1
large-v2
large-v3
large
large-v3-turbo
turbo
distil-small.en
distil-medium.en
distil-large-v2
distil-large-v3
distil-large-v3.5

The distil-* entries are Hugging Face Distil-Whisper models, not original OpenAI checkpoints.

whisper.cpp is a C/C++ runtime. Its downloadable ggml model list includes both normal and quantized files:

tiny
tiny.en
tiny-q5_1
tiny.en-q5_1
tiny-q8_0

base
base.en
base-q5_1
base.en-q5_1
base-q8_0

small
small.en
small.en-tdrz
small-q5_1
small.en-q5_1
small-q8_0

medium
medium.en
medium-q5_0
medium.en-q5_0
medium-q8_0

large-v1

large-v2
large-v2-q5_0
large-v2-q8_0

large-v3
large-v3-q5_0

large-v3-turbo
large-v3-turbo-q5_0
large-v3-turbo-q8_0

That is 30 whisper.cpp downloadable file variants, but most are just quantized versions of the same underlying models. In the script, .en means English-only, q5_* / q8_0 mean quantized, and tdrz means TinyDiarize.

Simple understanding

Model = tiny/base/small/medium/large-v1/large-v2/large-v3/large-v3-turbo
Language suffix = .en only for tiny/base/small/medium
Runtime = openai-whisper / faster-whisper / whisper.cpp / transformers
Format = PyTorch / CTranslate2 / ggml / ONNX / Core ML / MLX
Compression = fp32 / fp16 / int8 / q5 / q8

Best accuracy:        large-v3
Best speed/quality:   large-v3-turbo or turbo
Best English small:   small.en or medium.en
Best translation:     large-v3, large-v2, large, medium
Avoid for translation: turbo
Best Mac local path:  whisper.cpp + large-v3-turbo or large-v3
Best Python server:   faster-whisper + large-v3-turbo

For additional context, see AI audio transcription.