MiniMaxMiniMax-M2
MiniMax-M2 is a MiniMax open-weights reasoning model, released 27 Oct 2025. Its best configuration ranks #356 of 785 on the BenchLeader Index at 50.8 ±4.6, in the upper half. The ± is the point: 312 other configurations score within that range, so they and this one cannot be told apart on quality alone — price and speed are what separate them. It scores highest in instruction following (71) and lowest in knowledge (41). At $0.525 per million tokens blended it is mid-priced. Output speed of 99 tokens per second puts it faster than most, with a first answer in 21.6 s. Last measured 23 Oct 2026.
- Blended price
- $0.525/M
- $0.300 in · $1.20 out
- Output speed
- 99 tok/s
- measured by Artificial Analysis
- First answer
- 22 s
- first token 0.91 s
- Context
- 205k
- Full answer
- 27 s
- median, reasoning included
- Cost per run
- –
- Released
- 27 Oct 2025
- Overall index51
- Reasoning47
- Coding45
- Agents & tools45
- Maths52
- Knowledge41
- Instruction following71
- Human preference51
- Long context57
- Composite50
How its index has moved
8 Sept 2026 to 23 Oct 2026The index is recomputed from scratch every day, so a line moves when a new benchmark result lands, when a publisher revises a score, or when the models it is normalised against change. Early movement usually means the score is still settling.
Benchmark results
See all LLM benchmarks. One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.
Reasoning
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Arena Hard Prompts | 1367#199 | Arena (LMArena) | |
| GPQA Diamond (AA)not in index | 77.7%#217 | Artificial Analysis | |
| Humanity's Last Exam (AA)not in index | 13.7%#269 | Artificial Analysis | |
| Kagi LLM Benchmark | 57.8%#57 | Kagi LLM Benchmark | |
| CritPt | 0.9%#237 | Artificial Analysis |
Coding
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Arena Coding | 1383#212 | Arena (LMArena) | |
| Arena WebDev | 1297#119 | Arena (LMArena) | |
| SWE-bench Verified (bash only) | 61.0%#22 | SWE-bench | |
| SWE-bench Verified (any scaffold)not in index | 61.0%#27 | SWE-bench |
Agents & tools
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Terminal-Bench | 30.0%#60 | Terminal-Bench | |
| Terminal-Bench Hard | 25.8%#135 | Artificial Analysis | |
| τ²-Bench Telecom (AA)not in index | 86.8%#83 | Artificial Analysis | |
| Vending-Bench 2not in index | 160.6#62 | Andon Labs |
Maths
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Arena Maths | 1352#197 | Arena (LMArena) |
Knowledge
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| AA-Omniscience | -46.5#362 | Artificial Analysis | |
| AA-Omniscience: accuracynot in index | 23.2%#261 | Artificial Analysis | |
| AA-Omniscience: non-hallucinationnot in index | 9.1%#438 | Artificial Analysis |
Instruction following
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| IFBench | 72.3%#49 | Artificial Analysis | |
| Arena Instruction Followingnot in index | 1333#203 | Arena (LMArena) |
Human preference
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Arena Text | 1344#207 | Arena (LMArena) | |
| Arena Creative Writingnot in index | 1284#234 | Arena (LMArena) | |
| Arena Multi-turnnot in index | 1358#187 | Arena (LMArena) | |
| Arena Longer Queriesnot in index | 1339#216 | Arena (LMArena) |
Long context
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| AA-LCR | 64.3%#257 | Artificial Analysis |
Composite
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| AA Intelligence Index v4.3.2 | 18.6#249 | Artificial Analysis |
Where to run it
Every provider serving this model through OpenRouter, with throughput and first-token latency measured on live traffic over the last 30 minutes and each provider’s own price. Purple marks the best in each column.
| Provider | Speed | First token | Input $/M | Output $/M | Context | Quantisation |
|---|---|---|---|---|---|---|
| MiniMax | 51 tok/s | 0.91 s | $0.255 | $1.02 | 205k | fp8 |
What a task costs
Estimates from list price, output speed and time to first answer for the best configuration. “With caching” assumes three-quarters of the input is served from the prompt cache. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0005 | – | 24.6 s |
| Summarize a 30-page report | 12,000 / 600 | $0.0043 | – | 27.6 s |
| Code edit | 6,000 / 1,500 | $0.0036 | – | 36.7 s |
| Agentic coding session | 60,000 / 4,000 | $0.023 | – | 1.0 min |
| Structured extraction | 2,000 / 200 | $0.0008 | – | 23.6 s |
See also
Questions people ask
- What are MiniMax-M2's benchmark scores?
- As of 25 Oct 2026, MiniMax-M2 scores 50.8 ±4.6 on the BenchLeader Index, strongest in instruction following (71) and weakest in knowledge (41). That rests on 14 independent benchmark results, and every score on this page links to its publisher.
- How does MiniMax-M2 rank?
- MiniMax-M2 is #356 of 785 ranked configurations on 25 Oct 2026. 312 other configurations score within its margin, so the exact position matters less than the band. The leader is Claude Opus 5.5 (max effort) at 72.6.
- How much does MiniMax-M2 cost per 1M tokens?
- As of 25 Oct 2026, MiniMax-M2 costs $0.300 per 1M input tokens and $1.20 per 1M output tokens, $0.525 blended (three input tokens per output token).
- When was MiniMax-M2 released?
- MiniMax released MiniMax-M2 on 27 Oct 2025.
Data as of 25 Oct 2026. Compare with another model.
Cite as: BenchLeader, “MiniMax-M2: benchmarks, pricing, speed and rank”, data as of 25 Oct 2026, https://www.benchleader.com/models/minimax-m2.