Mistral Large 3

Mistral · Mistral Large · frontier · released Dec 2, 2025 · mistral-large-2512

Mistral's open-weight multimodal flagship with unusually low API pricing relative to its class.

Pricing

Input$0.5 / 1M

Cached inputNot surfaced in model page

Output$1.5 / 1M

Open-weight model with hosted API pricing listed on the model page.

Cost bandBalanced

Free tierOpen weights; hosted inference is paid.

Context window256k checked Jul 19, 2026

Max outputSee docs

Knowledge cutoffNot surfaced in summary

DeploymentMistral API or self-hosted open weights

OpenAI-compatible APIYes

ModalitiesText, Image input

Best forTeams that want a modern flagship model with open deployment paths and cheap hosted inference.

Business fitUnusually compelling if you want optional self-hosting without giving up mainstream capability.

Function callingAgentsBuilt-in toolsStructured outputsOCRDocument QnAFIM

Composite80 / 100

Coding82 / 100 (HumanEval)

Reasoning78 / 100 (MMLU-Pro)

Long context75 / 100 (RULER 128K)

Vision76 / 100 (MMMU)

Instruction following82 / 100 (MT-Bench)

Editorial aggregate of published benchmark results as of Dec 1, 2025. These are not independent measurements run by this site.

Dec 2, 2025New modelMistral Large 3 released as open weights Multimodal flagship at $0.5/$1.5 per 1M hosted, plus full self-hosted deployment path.

Last checkedJul 19, 2026

Cross-checked againstopenrouter.ai/mistralai/mistral-large-2512

Verify with provider documentation before committing spend. Report an error and it gets fixed with a logged correction.