Mistral Medium 3.5
Mistral AI · released Apr 30, 2026
Mistral Medium 3.5 is a dense 128B instruction-following model from Mistral AI. It supports text and image inputs with text output, and is designed for agentic workflows, coding, and complex...
Specification
- Context window
- 262K
- Max output
- —
- Knowledge cutoff
- Not stated
- Parameters
- Undisclosed
- Licence
- Proprietary
- Serving providers
- 1
- Moderated
- No
- Uptime
- 99.6%
Intelligence
30.4
74th percentile
Coding
46.9
Coding Index
Agentic
19.2
Agentic Index
Output speed
22 t/s
Median across providers
Latency
4s
Time to first token
Cost per task
$0.68
Estimated
Benchmarks
Where the score comes from
The Intelligence Index is a composite. These are the underlying evaluations this model was actually measured on.
Evaluation scores
Percentage correct · higher is better
- τ²-bench (Telecom)94.2%
- GPQA Diamond74.8%
- IFBench68.8%
- AA-LCR (long context)65.3%
- SciCode39.6%
- Terminal-Bench Hard33.3%
- Humanity's Last Exam13.8%
An evaluation missing from this list was not run for this model — it is not a zero.
View as table
| Evaluation | Score |
|---|---|
| τ²-bench (Telecom) | 94.2% |
| GPQA Diamond | 74.8% |
| IFBench | 68.8% |
| AA-LCR (long context) | 65.3% |
| SciCode | 39.6% |
| Terminal-Bench Hard | 33.3% |
| Humanity's Last Exam | 13.8% |
Against its peers
Intelligence Index · this model highlighted, nearest peers in grey
- GPT-5.1-Codex-MiniOpenAI31.3
- Ring-2.6-1TInclusionAI31.1
- o3OpenAI31.1
- Step 3.7 FlashStepFun30.9
- Mistral Medium 3.5Mistral AI30.4
- Qwen3.5-35B-A3BQwen29.9
- Claude Sonnet 4.5Anthropic29.9
- MiniMax M2MiniMax28.9
Peers are the models sitting closest on the Intelligence Index — the set you would realistically choose between.
View as table
| Model | Intelligence |
|---|---|
| GPT-5.1-Codex-Mini | 31.3 |
| Ring-2.6-1T | 31.1 |
| o3 | 31.1 |
| Step 3.7 Flash | 30.9 |
| Mistral Medium 3.5 | 30.4 |
| Qwen3.5-35B-A3B | 29.9 |
| Claude Sonnet 4.5 | 29.9 |
| MiniMax M2 | 28.9 |
Percentile among all indexed models
Pricing
What it costs to run
List prices per million tokens, plus what one representative task works out to.
List price
- Input / 1M tokens
- $1.5
- Output / 1M tokens
- $7.5
- Cached input / 1M
- Not offered
- Blended 3:1
- $3
One task, estimated
$0.68
- Input tokens
- 50,000
- Output tokens
- 80,000
- Profile
- Reasoning
Estimated from list pricing: 50K input tokens plus 80K output tokens for reasoning models (25K for non-reasoning).
Serving providers
Speed and latency figures are medians across these providers, so a widely-served model reports a blend rather than any single endpoint.
Arena
Head-to-head generation quality
Elo from pairwise judgements, broken out by the kind of thing the model was asked to build.
- Overall Elo
- 1,164
- Win rate
- 42.4%
- Strongest at
- Websites
- Tournaments
- 5,837
Elo by category
Dot position on a 975–1,175 scale · Elo has no meaningful zero
- 1,164
- 1,148
- 1,121
- 1,115
- 1,114
- 994
Agent categories (full-stack apps, mobile apps) are only scored for models tested in agent harnesses.
View as table
| Category | Elo | Win rate |
|---|---|---|
| Websites | 1164 | 42.4% |
| Data visualisation | 1148 | 43.2% |
| Game development | 1121 | 38.6% |
| 3D scenes | 1115 | 38.3% |
| UI components | 1114 | 37.3% |
| SVG | 994 | 26.6% |