Ministral 3
Ministral 3
A three-billion-parameter model that answers almost before you finish the request.
Updated
The read
Ministral is Mistral's edge tier, and speed is the entire pitch: among the highest measured output rates of any hosted model, at a size small enough to sit on a device rather than in a datacentre. Use it for autocomplete, intent detection, routing and anything else where a hundred milliseconds is the difference between useful and annoying. It knows very little, and that is by design.
Where it shines
- Among the fastest models measured anywhere
- Small enough for on-device and edge deployment
- Cheap enough to call on every keystroke
Worth knowing first
- Very limited world knowledge and reasoning
- Not published under a permissive open licence
Index score
69.0
Provisional. This score is read from the product's public capability, not from the one-prompt rebuild the fully reviewed entries went through, so treat it as a placing rather than a verdict.
- Rank
- 70 of 73
- Builds
- LLMs
- Category
- Small and fast
- Output
- Text over an API, sized for the edge
- Pricing
- Paid only
- Best for
- Latency-critical work like autocomplete and routing
Scores are our own editorial judgement, weighted craft 55, speed 10, control 15 and value 20.
Also worth a look
Full index- 11Claude Haiku 5.5Claude Haiku 5.5ProvisionalA small model priced like a rounding error that scores where mid-tier models did last quarter.Small and fastPaid only87.9
- 22GPT-6 LunaGPT-6 LunaProvisionalThe cheap end of GPT-6, and the model free ChatGPT users now talk to.Small and fastPaid only84.9
- 30GPT-5.6 LunaGPT-5.6 LunaProvisionalStill cheap and still capable, but GPT-6 Luna costs half as much.Small and fastPaid only83.0
