Capabilities & modalities
Mistral's open-weight models are strong generalist, coding and multilingual text models, and
Pixtral adds native image input. They are solid general assistants with a normal hallucination
profile for their size. Per aggregated leaderboards (ev-leaderboard, ev-small3) they are
competitive to strong within each size class. Mistral 7B and Mixtral were class-leading at
release, and Mistral Small 3.x is competitive for a 24B while adding tool use and structured
output. Mistral suits a deployer who wants a strong, EU-domiciled open model with Apache
licensing on the flagship checkpoints.
Context window & long-context behaviour
Context length varies by model, so confirm the number on the model card. Mistral Nemo (12B)
and Mistral Small 3.x (24B) offer 128K tokens, Mixtral 8x7B and 8x22B roughly 32k-64k, and
Mistral 7B 8k-32k depending on version. The published maximum is the architectural ceiling, so
validate quality at the target length.
Use the model's chat template instead of hand-rolling role markers. mistral-common is the
ground truth for tokenization and templating, newer models use the Tekken tokenizer
(tiktoken-based, larger vocabulary), and the Hugging Face chat template matches
mistral-common output.
messages = [
{"role": "system", "content": "You are a helpful assistant."},
{"role": "user", "content": "Summarize the following..."},
]
prompt = tok.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
Base checkpoints have no chat template. They are plain text-completion models and should not
be prompted as chat assistants. A wrong template or wrong special tokens is the most common
cause of garbled or over-verbose output.
Language coverage
Mistral is multilingual, with particular strength in European languages (English, French,
German, Italian, Spanish and more), a natural fit for its EU market. Even the companion
Moderation API spans 11 languages (ev-moderation). Validate per-language quality for
production, especially for lower-resource languages, because per-model coverage varies.
The instruct models support native function and tool calling, a documented capability of the
flagship releases. Mistral Small 3.2 improved tool-use accuracy over 3.1 (ev-small3). Use
the model's tool-call schema via the chat template or mistral-common instead of improvising
a prompt format, and validate the emitted call structure, since some releases note residual
variability in tool-use reliability.
Structured / JSON-constrained output
The instruct models support JSON and structured output, and Mistral Small 3.2 improved
structured-output quality over 3.1 (ev-small3). The item is marked partial because hard
schema-constrained decoding, a guarantee that output conforms to a grammar or JSON schema, is
delivered at the serving layer (guided decoding or grammars in vLLM, TGI, llama.cpp) and not
as a model-level guarantee. Without constrained decoding, structured output from the model is
strong but best-effort.
The ownership factor this domain covers, drawn from the one entry record.
3
ReliabilityIs it reliable and good enough for the job?
ModerateThe models run dependably across every major runtime. Safety tuning is lighter than at the largest US labs, so pair the instruct variant with the Moderation API and your own guardrails.
How this scores (AOI sub-dimensions)
Operational4/5how practical it is to run, serve and maintain in productionEcosystem support is broad: an official inference library (mistral-inference), day-0 or near-day-0 loading on vLLM, llama.cpp, Ollama, TGI and transformers, safetensors distribution, community quants and hosted endpoints on the major clouds.
Safety3/5whether misuse risks are evaluated and guardrails are providedInstruct variants are safety-tuned, and Mistral ships a hosted Moderation API (a fine-tuned Ministral 8B) and system-level guardrailing guidance, which is above a bare release.
The same evidence records as the entry sheet. Read means the text was verified; unverified means it is known to exist but not yet read.
Documentationread2026-07-25
Mistral's model-weights documentation page assigns licences per model.
Model cardunverified2026-07-25
Mistral open-weight checkpoints are hosted on the verified mistralai organisation on Hugging Face in safetensors with per-file checksums.
Vendor announcementunverified2026-07-25
Mistral Small 3 (24B) is released under Apache-2.0 as Mistral recommits to Apache for general-purpose models and moves away from MRL; it offers a 128K context window, tool use and structured output.
Licenceread2026-11-19
The Mistral Research Licence (MRL-0.1) permits use only for "Research Purposes".
Licenceread2026-11-19
The Mistral AI Non-Production Licence (MNPL-0.1) restricts use: "You shall only use the Mistral Models and Derivatives (whether or not created by Mistral AI) for testing, research, Personal, or evaluation purposes in Non-Production Environments".
Vendor announcementunverified2026-07-25
Mistral Nemo (12B) is a 128K-context model released under Apache-2.0 in collaboration with NVIDIA, using the Tekken tokenizer.
Third-party analysisread2026-11-19
Mistral AI, a Paris-based EU company, is listed among the 22 signatories of the EU General-Purpose AI Code of Practice on the European Commission page.
Vendor announcementunverified2026-07-25
Mistral offers a hosted Moderation API - a fine-tuned Ministral 8B classifier spanning multiple harm categories and 11 languages - plus system-level guardrailing guidance for downstream deployments.
Documentationunverified2026-07-25
Mistral documents its instruct tokenization and chat templates: mistral-common is the ground truth, newer models use the Tekken tokenizer, and the Hugging Face chat template matches mistral-common output.
Documentationunverified2026-07-25
mistral-inference is Mistral's official inference library; the models also load on vLLM, llama.cpp, Ollama, TGI and transformers.
Third-party analysisunverified2026-07-25
Mistral models are competitive-to-strong within their size classes on independent public leaderboards.
Third-party analysisunverified2026-07-25
Mistral AI is a France-based lab whose open-weight lineage (Mistral 7B, Mixtral, Nemo, Mistral Small, specialist variants) is broadly adopted, with mixed Apache-2.0 and research-only licensing and wide ecosystem/cloud availability.
Privacy Policyread2026-11-19
Mistral's Privacy Policy (effective 3 September 2026) covers non-EU service providers with GDPR Article 46 safeguards, with the European Commission's Standard Contractual Clauses attached.
Documentationread2026-11-19
Per the Mistral help centre, on Vibe Free, Pro, Teams and Student, input and output are used by default to train models unless the user opts out.
Privacy Policyread2026-11-19
Mistral Privacy Policy section 5 (effective 3 September 2026): for the API, Input and Output are kept for the period necessary to generate the Output and then for thirty (30) rolling days to monitor abuse (unless zero data retention is activated).
Documentationread2026-11-19
Mistral help centre: Zero Data Retention "is available only with pay-as-you-go and only for stateless API calls".
Data Processing Addendumread2026-11-19
Mistral's Data Processing Addendum (effective 27 July 2026) lets Mistral process Personal Data as Controller to train its models "unless Customer is or has opted-out of training" and for abuse monitoring except under zero data retention (2.3).
Model cardread2026-11-19
On the raw Hugging Face model cards and API, Mistral-Large-3-675B-Instruct-2512 and -Base-2512, Ministral-3-3B/8B/14B and Mistral-Small-4-119B-2603 carry license apache-2.0 in the card front matter.
Documentationread2026-11-19
The Mistral docs Models Overview labels Mistral Small 4, Ministral 3 and Voxtral Small as Apache 2.0, Mistral Large 4 as Mistral Ecosystem Open Weight License and Timestral as Modified MIT.
Terms of serviceread2026-11-19
Mistral Commercial Terms of Service (effective 25 September 2026): Mistral will not use Customer Data or Outputs to train its models except in four cases (clause 4.2).