M2.1
MiniMax · released Dec 23, 2025
MiniMax-M2.1 is a lightweight, state-of-the-art large language model optimized for coding, agentic workflows, and modern application development. With only 10 billion activated parameters, it delivers a major jump in real-world...
Specification
- Context window
- 205K
- Max output
- 131K
- Knowledge cutoff
- Not stated
- Parameters
- 228.7B
- Licence
- other
- Serving providers
- 2
- Moderated
- No
- Uptime
- 100.0%
Intelligence
32.1
77th percentile
Coding
—
Coding Index
Agentic
—
Agentic Index
Output speed
66 t/s
Median across providers
Latency
668ms
Time to first token
Cost per task
$0.11
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
- MMLU-Pro87.5%
- τ²-bench (Telecom)85.4%
- GPQA Diamond83.0%
- AIME 202582.7%
- LiveCodeBench81.0%
- IFBench69.9%
- AA-LCR (long context)66.3%
- SciCode40.7%
- Terminal-Bench Hard28.8%
- Humanity's Last Exam23.2%
An evaluation missing from this list was not run for this model — it is not a zero.
View as table
| Evaluation | Score |
|---|---|
| MMLU-Pro | 87.5% |
| τ²-bench (Telecom) | 85.4% |
| GPQA Diamond | 83.0% |
| AIME 2025 | 82.7% |
| LiveCodeBench | 81.0% |
| IFBench | 69.9% |
| AA-LCR (long context) | 66.3% |
| SciCode | 40.7% |
| Terminal-Bench Hard | 28.8% |
| Humanity's Last Exam | 23.2% |
Against its peers
Intelligence Index · this model highlighted, nearest peers in grey
- o3 ProOpenAI33.3
- 32.8
- 32.5
- M2.1MiniMax32.1
- Qwen3.6 35B A3BQwen32.1
- GPT-5.1-Codex-MiniOpenAI31.3
- Ring-2.6-1TInclusionAI31.1
- o3OpenAI31.1
Peers are the models sitting closest on the Intelligence Index — the set you would realistically choose between.
View as table
| Model | Intelligence |
|---|---|
| o3 Pro | 33.3 |
| Qwen3.5-122B-A10B | 32.8 |
| Qwen3 Max Thinking | 32.5 |
| M2.1 | 32.1 |
| Qwen3.6 35B A3B | 32.1 |
| GPT-5.1-Codex-Mini | 31.3 |
| Ring-2.6-1T | 31.1 |
| o3 | 31.1 |
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
- $0.3
- Output / 1M tokens
- $1.2
- Cached input / 1M
- $0.03
- Blended 3:1
- $0.525
One task, estimated
$0.11
- 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).
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
- 1,258
- Win rate
- 60.9%
- Strongest at
- UI components
- Tournaments
- 746
Elo by category
Dot position on a 1,175–1,275 scale · Elo has no meaningful zero
- 1,258
- 1,233
- 1,224
- 1,220
- 1,178
- 1,176
Agent categories (full-stack apps, mobile apps) are only scored for models tested in agent harnesses.
View as table
| Category | Elo | Win rate |
|---|---|---|
| UI components | 1258 | 60.9% |
| Data visualisation | 1233 | 57.0% |
| Websites | 1224 | 55.4% |
| 3D scenes | 1220 | 57.5% |
| SVG | 1178 | 55.4% |
| Game development | 1176 | 50.4% |