Mistral Vibe
Mistral's AI chat assistant and agent for work and code, formerly branded Le Chat.
Core record
Status: Public
Access: Web / Mobile / CLI / Editor
API status: Via Mistral platform
Openness: Mixed
Modalities: Text, code, voice and connected tools
Best suited for
Work, coding, connected-tool and agent workflows in the Mistral ecosystem.
Limitations
Vibe product features and the underlying Mistral model lineup can evolve independently.
API / auth
Interface: Mistral API
Base URL: https://api.mistral.ai/v1
Authentication: Bearer Mistral API key
Compatibility: Native Mistral API with chat/completions and agents/conversations endpoints.
Rate limits: Subscription tier and model dependent.
Deployment
Models / versions
| Model | Context | Max output | Modalities | License | Status |
|---|---|---|---|---|---|
| Mistral Medium 3.5 mistral-medium-3-5 | 256K tokens | Pending | text, multimodal | Modified MIT | GA |
| Mistral Small 4 | 256K tokens | Pending | text, multimodal | Apache 2.0 | current |
| Mistral Large 3 | 256K tokens | Pending | text, multimodal | Apache 2.0 | current |
Pricing snapshots
| SKU / model | Unit | Input | Output | Flat | Notes |
|---|---|---|---|---|---|
| Mistral Medium 3.5 | USD per 1M tokens | 1.5 | 7.5 | — | |
| Mistral Small 4 | USD per 1M tokens | 0.15 | 0.6 | — | |
| Mistral Large 3 | USD per 1M tokens | 0.5 | 1.5 | — |
Integrations
Regions
Mistral offers regional inference for eligible workloads.
Governance / privacy
Mistral publishes a lifecycle policy and explicit model deprecation/replacement information.
Licenses vary by model; openness must be stored at model-version level.
Benchmarks
Pending methodology-aware benchmark audit unless separately documented. Vendor and independent benchmark results are kept distinct.