What is Mistral Le Chat
Mistral’s assistant product, previously known as Le Chat and now branded as Vibe, is increasingly interesting because it combines a polished assistant experience with Mistral’s credibility in open and commercial model development. It is designed for both productivity work and coding, rather than only casual chat.
Mistral rose quickly from a European frontier-model startup into one of the most important alternative AI vendors for teams that care about open-weight strategy, deployment flexibility, and strong technical products. Its assistant surface evolved from Le Chat into Vibe, reflecting the company’s shift from model reputation alone toward a broader platform and workflow story.
Core offerings
- Vibe for work and Vibe for code inside one product.
- Web, mobile, terminal, IDE, and connector-based workflows.
- Studio for API and developer access plus Admin for org controls.
- 100+ connectors and options for private or on-prem deployment on Enterprise.
Mistral now presents its assistant and coding product as Vibe across web, CLI, IDE, and mobile, with Fast and Think controls, connectors, tasks, and team administration. Enterprise customers can also discuss private or on-premises deployment with sales.
Pricing
Mistral’s public pricing currently lists Free, Pro at US$14.99 per month, Team at US$24.99 per user per month, and Enterprise with custom pricing. Verified students can access Pro for US$5.99 per month for up to 12 months when eligible. The current plan comparison lists US$25.50 per month in API credits for Pro and US$15 per month for the Education plan; it does not list a monthly API-credit allowance for Free. Mistral’s current API example prices Mistral Large at US$0.50 per million input tokens and US$1.50 per million output tokens; batch requests are 50% cheaper and cached input can be discounted by up to 90%.
Model footprint
Mistral exposes multiple products and model routes rather than a single assistant model. Its current public catalogue highlights Mistral Large 3, Medium 3.5, Small 4, OCR 4, and Voxtral text-to-speech alongside specialised routes for coding and document work.
Why select Mistral Le Chat
Mistral appeals to teams that want a modern assistant plus stronger infrastructure and deployment flexibility than many consumer-first AI products provide. It is particularly interesting for Europe-conscious buyers, technical teams, and companies that care about open-weight strategy.
Official sources: Mistral pricing, Mistral platform overview, Mistral Vibe, Mistral release notes.
Current models
Mistral’s public docs are strong on model families, open-weight status, context windows, and pricing for current flagship lines. Training data details are still limited publicly, so those cells remain explicit about that.
| Model | Context / output | Knowledge / training | Pricing / token use | Speed / notes |
|---|---|---|---|---|
Mistral Large 3mistral-large-2512 | 256K max context | Open-weight, Apache 2.0 Parameters: 675B total / 41B active Training data specifics: not publicly disclosed | US$0.5 input / US$1.5 output per 1M tokens | State-of-the-art general-purpose multimodal flagship with unusually low published token pricing |
Mistral Medium 3.5mistral-medium-3-5 | 256K max context | Modified MIT licence Parameters: 128B total / 128B active Training data specifics: not publicly disclosed | US$1.5 input / US$7.5 output per 1M tokens | Frontier-class multimodal model optimised for coding and agentic use |
Mistral Medium 3.1mistral-medium-2508 | 128K context | Premier release from Aug 2025 Training data specifics: not publicly disclosed | Current consumer/platform pricing pages emphasise newer lines; check the live model card for support lifecycle | Still relevant for understanding how fast Mistral iterates its mid-tier model family |
Magistral Medium 1.2magistral-medium-2509 | 128K max context | Reasoning-focused line Training data specifics: not publicly disclosed | Price not exposed in the snippet used here; see current compare page for live values | Useful when you specifically want Mistral’s reasoning-oriented family rather than generalist models |
Ministral 3 familyministral-3b-2512, ministral-8b-2512, ministral-14b-2512 | 256K max context across the line | Open model family focused on smaller footprints Training data specifics: not publicly disclosed | Current pricing varies by deployment route rather than one simple Vibe plan row | Important for teams that care about smaller open models and deployment flexibility |