MiniMax
MiniMax: MiniMax M2.7
minimax/minimax-m2.7
MiniMax-M2.7 is a next-generation large language model designed for autonomous, real-world productivity and continuous improvement.
- Context window: 204,800 tokens
- Input: text
- Output: text
- Pricing: $0.21/M input tokens, $0.84/M output tokens
Community sentiment
Negative (-18) — 3 positive, 6 negative, 1 neutral across 10 community samples from X and the web, as of 2026-09-22.
Users highlight MiniMax M2.7’s reasoning, coding and native multi-agent collaboration and report easy/free access with low observed latency in quick tests. Criticisms focus on efficiency and benchmark gaps versus newer models (higher active-parameter cost for similar scores), lower throughput/context compared with peers, and occasional inconsistency in agent runs.
- “MiniMax M2.7 brings another interesting capability to AINFT’s model ecosystem. Its documentation highlights reasoning, coding and native multi-agent collaboration. That is important because multi-agent systems require models that can participate effectively in coordinated https:/” — X
- “how to run deepseek v4 flash and minimax m2.7 for $0 😳 routeway keeps a few models on free ids, so running them never touches your balance 1. make an account at https://t.co/wcmBHrnicO 2. create an api key 3. paste https://t.co/tLbHbxaDcV into cherry studio or any https://t.co/y” — X
- “Context is ~256K. It’s verbose though — ~67k output tokens per Intelligence Index task (~24k answer + ~43k reasoning), about 3.2× MiniMax-M2.7. Speed: ~159 output tok/s in its size class. Takeaway: finance-tuned open MoE that’s free to call and efficient on active params. https:/” — X
- “Luna is why I bothered with repeated runs. One run scored 100/100. The other two failed. Terra and MiniMax M2.7 High Speed both finished 2/3. A clean demo tells you very little about whether an agent will get through the job tomorrow. https://t.co/DipgFzRP1d” — X
View on OpenRouter.
Model data sourced from OpenRouter.