MiniMax M2.7
minimax/minimax-m2.7- MiniMax
- Family · MiniMax M2
- Released 18 Mar 2026
- Open weights
- Text → Text
- Tools
Not on OSMEXO yet
Spec & list price
Context200K
Max output128K
Input / 1M$0.30
Output / 1M$1.204× input
Scores · standing
Intelligence22.8#113/490
Coding52.6#61/196
Usage
Tokens / day2.0B▼ 36% 7d
Usage rank#158
Requests / day108.2K
Tool calls / req0.26
Per request · 30 days
Avg input17.4K1.2× median
Avg output5780.8× median
Cache hit71%about median
Reasoning share53%about median
Tool-call failures4%2.1× median
Consumption
Peak 11.1B 30 SepLow 1.2B 1 NovRank today #158
News & trends
News · latest from MiniMax
Benchmarks
bar = share of the top score · Benchmark data: Artificial AnalysisSpeed · median across 6 providers, latest day
Output speed#261/33443 tok/s1.5× slower than median
Fastest provider307 tok/sbest of 6 providers
First token#2071.37 s1.3× longer than median
#9 cheapest of 32 with a similar score
Value
Intelligence Index
Blended price per 1M tokens (log)
Family progress
Intelligence Index
- Family releases22.8
Technical details
- Tokenizer
- Other
- License
- other
- Parameters
- 229B
- Moderation
- none
- Downloads · likes
- 330.3K · 1.3K
Request parameters 16
- tools
- tool_choice
- response_format
- reasoning
- include_reasoning
- temperature
- top_p
- top_k
- min_p
- max_tokens
+6 more
- stop
- seed
- frequency_penalty
- logit_bias
- presence_penalty
- repetition_penalty
Interesting facts
- The most used of 4 MiniMax M2 models by tokens last month.
- Released 229 days ago, the newest of 4 MiniMax M2 models.
When to use
Best for
- Instruction-following assistants
- Tool-calling agents with long context
- Science and technical question answering
- Self-hosted deployments (open weights)
Think twice for
- Image input (text only)
- Strict structured-output pipelines (not supported)
- Hardest coding agent work
Research papers
- SWE-VERIFIER: Selective Verification for Cost-Efficient Coding Agents21 Oct
- WavePP: High-Throughput Pipeline Parallel LLM Prefill under Prefix Reuse28 Sep
- Certified Long-Horizon Code Agent Evolution via Validation-Gated Skill Optimization26 Sep
- FinHardBench: Can LLMs Generate Latency-Aware Hardware for Financial Computing?2 Aug
- Benchmarking LLMs on File System Design and Implementation31 Jul
- Cross-Tokenizer On-Policy Distillation via Byte-Prefix Marginalization24 Jul
- ExpertPlex: A High-Goodput Disaggregated Serving System for MoE LLMs with Adaptive Persistent Kernels20 Jul
- EvoClawBench: Can Agents Learn Reusable Skills from Their Own Runs?23 Jun
Same family
MiniMax M2- 22.8
- 20.9
- 18.6
Alternatives
other vendors · similar capability- 26.5
- 26.3
- U26.6
- 21.8
Want MiniMax M2.7 on OSMEXO?
Sign in to request it. Requests are reviewed; a request does not guarantee availability.