MiniMax M2
MiniMax·Oct 2025reasoningopen weights · MIT
68.9
modhub Index
MiniMax M2 takes an opinionated position: optimize everything for agentic coding and tool-calling loops, keep only 10B parameters active for speed and cost, and ship the weights under MIT. The result was briefly the top open model on several agentic indices, at prices and latencies that make multi-agent fan-out architectures affordable. A favorite engine for high-concurrency agent swarms.
Benchmark results
Where it shines
- Exceptional efficiency — 10B active parameters
- Tuned specifically for agent/tool-call workloads
- MIT license and very low hosted pricing
Alternatives to MiniMax M2
OpenAI's dedicated 2025 reasoning model that pioneered thinking-with-images and agentic tool use within chain-of-thought.
The sparse-attention release that halved DeepSeek's already-lowest-in-class prices while keeping GPT-class quality.
A trillion-parameter open reasoning agent that can chain 200–300 tool calls — the open-weights agentic standout of late 2025.
OpenAI's August 2025 unified flagship that merged the GPT and o-series reasoning lines into one model with selectable effort.
Frequently asked questions
- How much does the MiniMax M2 API cost?
- MiniMax M2 costs $0.3 per million input tokens and $1.2 per million output tokens. A workload of 10M input and 1.5M output tokens per month costs about $4.80.
- What is the context window of MiniMax M2?
- MiniMax M2 supports a context window of 204,800 tokens (205K).
- Is MiniMax M2 open source?
- Yes — MiniMax M2 is an open-weights model released under the MIT license, so it can be downloaded and self-hosted.
- What are the best alternatives to MiniMax M2?
- The closest alternatives by overall capability are OpenAI o3, DeepSeek V3.2, Kimi K2 Thinking, GPT-5. See the comparison pages for detailed head-to-head breakdowns.