AI Model Release Tracker

Mistral Large 3Open Weight

Released

Mistral Large 3 is an AI model released by Mistral on Tuesday, Dec 2 2025, the same day as Ministral 3 14B-25-12. It is an open-weight model — the trained weights are available to download and run. It is a 675B parameterThe learned values that shape how a model responds. B means billion and T means trillion. For mixture-of-experts models, the total count is larger than the number active for each token. Parameter count alone does not measure quality. model with a 256k token context windowHow much information the model can work with in one request, measured in tokens — pieces of words or other data. This includes your question, attachments, conversation and reply. Limits can vary by model variant and provider.. Benchmark results (shown below) cover BullshitBench v2.

API pricing

InputWhat you pay for everything you send the model — your question, plus any documents or earlier conversation you include with it.
$0.50
Cached inputA reduced rate for text you send over and over. If every request starts with the same instructions or the same document, the provider keeps a copy ready and charges less to read it again.
$0.05
OutputWhat you pay for the text the model writes back. It is normally the dearer half: producing an answer costs more than reading one.
$1.50
USD per 1M tokensEvery price here is for one million tokens. A token is roughly three-quarters of a word, so a million tokens is about 750,000 words of text.Verified October 19, 2026Read from mistral.ai. Click the check to open the source.

Available from

Mistral—zdr$0.50$1.50256K
Mistraleu—$0.55$1.65256K
USD per 1M tokensEvery price here is for one million tokens. A token is roughly three-quarters of a word, so a million tokens is about 750,000 words of text.Checked October 19, 2026Third-party rates from openrouter.ai, not the lab’s own list price.

Benchmarks

BullshitBench v2Nonsense detection — Given a confidently-worded but nonsensical prompt, does the AI spot that it makes no sense and push back — instead of playing along and inventing an answer? The score is how often it clearly called out the nonsense. Higher is better.
2%
#96 of 97

Compare Mistral Large 3 with

Mistral Large 3

Suggested comparisons

Mistral Large 3vsMistral Large 4Mistral Large 3vsGPT-6.1 SolMistral Large 3vsClaude Haiku 5.5Mistral Large 3vsGemini 4 ArgonMistral Large 3vsMuse Spark OpenMistral Large 3vsGrok 4.7Mistral Large 3vsDeepSeek-V4.1-Flash-1017Mistral Large 3vsKimi K3Mistral Large 3vsGLM-5.3-FlashMistral Large 3vsQwen3.8-Max-0902Mistral Large 3vsNemotron 3.5 Super VL

About

Mistral Large 3 was released on December 2, 2025 as a sparse mixture-of-experts model with 675 billion total parameters and 41 billion active per token. It marked Mistral's return to a large sparse architecture after the Mixtral line. The release combined multilingual language capabilities with image understanding in a flagship model available as downloadable weights.

Mistral published both base and instruction-tuned versions under Apache 2.0, alongside compressed deployment formats and hosted access. The launch model was instruction-oriented; Mistral described a reasoning version as future work rather than a feature already supplied by that checkpoint. Its sparse computation reduced the active work per token, but serving still required infrastructure appropriate to the complete model and chosen precision.

Frequently asked questions

Mistral Large 3 was released by Mistral on Tuesday, Dec 2 2025.

All Mistral releases

33 tracked