Model · Mistral
Mistral
Efficient European models that work well as the model that merges the others.
Mistral builds lean, capable models that get a lot done per dollar. They are a natural fit for the synthesis step in a weave, where the job is to read several answers and combine them well.
The lineup
| Model | Best for | Plan |
|---|---|---|
| Mistral Medium 3.1 | The default free judge that merges model outputs. | Free |
| Mistral Large 2512 | A stronger Mistral model on the paid plans. | Paid |
What it is good for
- Synthesizing and judging several model answers
- Efficient, cost-effective general work
- A European-built perspective in the mix
In a weave
Mistral Medium 3.1 is the default judge in many weaves: the model that reads everyone’s answer and produces the merged result. It is free, which keeps the synthesis step cheap.
Mistral Medium 3.1 is included free and serves as the default judge. Mistral Large is on the paid plans.
Common questions
What does the judge model do?
In a weave, the judge reads the answers from the responding models and merges or ranks them into one result. Mistral Medium 3.1 is the free default for that role.
More models
Claude
Anthropic’s models. Careful reasoning, strong writing, and steady instruction-following.
OpenAIGPT
OpenAI’s models. Versatile all-rounders with strong general reasoning.
GoogleGemini
Google’s models. Huge context windows and fast, cheap Flash tiers.
DeepSeekDeepSeek
Open-weights models with strong reasoning at a low price, plus very long context.
AlibabaQwen
Alibaba’s models. Strong multilingual range, plus solid coding and math.
MetaLlama
Meta’s open-weights models. Efficient, compact, and widely adopted.
xAIGrok
xAI’s models. A distinct voice and conversational style.
Try Mistral in a weave
Run it next to other models on your own task. Free to start, no card.
Get started