Skip to content

Model catalogMistral

Mistral Large 3

Mistral's frontier model — bring your Hugging Face token. Served with vLLM behind your own private, OpenAI-compatible /v1 endpoint — billed by the second, terminate anytime.

675B MoE · 41B act · FP8 · 256K context · 8× 141 GB · per-second billing

01Specifications

The specs the launcher enforces.

Architecture, served precision, max context, the minimum GPU shape, and the exact pinned weights — the facts that decide how Mistral Large 3 runs and what it costs.

SpecValue
Architecture675B MoE · 41B act
Served precisionFP8
Max context256K
Minimum GPU shape8× 141 GB
Weights servedmistralai/Mistral-Large-3-675B-Instruct-2512@383ffea

Runs on 8× 141 GB · billed per second — final cost depends on the GPU shape.

Full pricing →

02How it runs

A private endpoint, not a shared API.

Launching Mistral Large 3 runs the model on your own instance and gives you a private, OpenAI-compatible /v1 endpoint and API key. Point any OpenAI-compatible SDK at your base URL, change the model name, and existing code works. The endpoint is yours alone — terminate the instance and it goes away with it.

Billing is by the second on the 8× 141 GB shape the launcher enforces, from boot to terminate. No commitment and no tier gates.

Gated model

Mistral Large 3 is a gated repository on Hugging Face. Accept the model's license on Hugging Face, and the launch flow injects your own token so the instance can pull the weights.

Limited GPU supply

8× 141 GB (H200/B200) instances are intermittently out of stock across providers. If the launch flow shows no compute for this shape, check back later — supply returns.

03Get started

Launch Mistral Large 3 now.

One private OpenAI-compatible endpoint, one key, billed by the second. The console opens with Mistral Large 3 preselected.