utWhat it does
Published capability labels: Capabilities unconfirmed. Listed uses: Research.
- Inputs
- See model card
- Outputs
- See model card
- Category
- Other
- Context window
- Not reported
- License
- other
These labels describe intended capabilities. Scout has not independently evaluated this model or established that it is best at a particular task.
Model size & local memory
Running requires extra memory for cache, activations, and the runtime. This is not a tested minimum VRAM requirement. A compatible 4-bit download is not guaranteed; mixture-of-experts estimates use total parameters.
No independently verified hardware speed is attached to this profile. You can record your own measurements in the explorer’s My hardware view. Understand the memory estimate.
Cloud pricing
No cloud price is linked to this listing. Cloud API entries are listed separately in the library.
Tested performance
Independent leaderboard results for exact configurations of this model. Each test has its own scale, so compare results within a test, not across tests. See every tested model.
Code Arena
Build apps & websites · Ratings published Oct 23, 2026
| Configuration | Result | 95% interval | Evidence |
|---|---|---|---|
| Mistral Medium 3.5 | 1264.2 | 1249.4 to 1279.0 | 2,315 votes · leaderboard rank #134 |
A preference rating, not a percentage of correct answers. Design, usability, and the coding setup affect votes. Original leaderboard ↗ · CC BY 4.0
Agent Arena
Work with an AI agent · Ratings published Oct 23, 2026
| Configuration | Result | 95% interval | Evidence |
|---|---|---|---|
| Mistral Medium 3.5 | -0.121 | -0.139 to -0.104 | 13,468 sessions · leaderboard rank #45 |
A net improvement index relative to Arena’s baseline. Zero is the baseline, not zero capability; this is not a task completion percentage. Original leaderboard ↗ · CC BY 4.0
Benchmark evidence
No vendor-reported results are attached to this listing.