MiniMax M2 / M2.5 Pricing
All plans and pricing as of 2026-04-13
Free tier available2 plansExcellent value score (9.5/10)
Self-hosted (Free)
$0
- ✓MIT license on M2 and M2.5
- ✓Weights on Hugging Face
- ✓Commercial use permitted
Most Popular
API (OpenRouter / MiniMax)
$0.30/per 1M input tokens
- ✓M2: $0.30 in / $1.20 out
- ✓192K+ context
- ✓Native agentic + tool-use
Is MiniMax M2 / M2.5 Worth the Price?
S
Value Score: 9.5/10
Overall Score: 8.4/10 · Agentic coding and tool-use workflows on a budget. Best price-to-SWE-Bench ratio of any open-weights model in 2026.
MiniMax M2/M2.5 is the most cost-efficient frontier-tier open model in 2026. The 80.2% SWE-Bench Verified score is a genuine breakthrough -- matching Claude Opus 4.6 on real coding tasks at a tenth of the price. The sparse 10B-active MoE runs fast on moderate hardware. The main drawback is ecosystem: MiniMax has less Western infrastructure support than Alibaba or DeepSeek. If you're building an agentic product and want maximum value per token, M2.5 is an A-tier pick.
How MiniMax M2 / M2.5 Pricing Compares
| Tool | Free Tier | Starting Price | Value Score | Overall |
|---|---|---|---|---|
| MiniMax M2 / M2.5(this tool) | Yes | $0 | 9.5/10 | 8.4 |
| Qwen (Alibaba) | Yes | $0 | 10/10 | 8.8 |
| Gemma 4 (Google) | Yes | $0 | 10/10 | 8.3 |
| Kimi K2.5 (Moonshot) | Yes | $0 | 8.5/10 | 8.1 |
| DeepSeek | Yes | $0 | 9.5/10 | 8.0 |
| GLM / Z.ai (Zhipu AI) | Yes | $0 | 9/10 | 8.0 |