MiniMax-M2.5
MiniMax's flagship model matching frontier performance at 1/20th the cost with 80.2% SWE-bench Verified
MiniMax-M2.5
MiniMax • February 2026
Training Data
Up to early 2026
MiniMax-M2.5
February 2026
Parameters
230B (10B active)
Training Method
Mixture of Experts with RL
Context Window
200,000 tokens
Knowledge Cutoff
January 2026
Key Features
Frontier Performance • Ultra Low Cost • Open Weights (MIT) • Agentic Workflows
Capabilities
Coding: Outstanding (80.2% SWE-bench)
Search: Outstanding (76.3% BrowseComp)
Tool Use: Excellent
What's New in This Version
Matches Claude Opus 4.6 and GPT-5.2 on key benchmarks at 10-20x lower cost
MiniMax's flagship model matching frontier performance at 1/20th the cost with 80.2% SWE-bench Verified
What's New in This Version
Matches Claude Opus 4.6 and GPT-5.2 on key benchmarks at 10-20x lower cost
Technical Specifications
Key Features
Capabilities
Other MiniMax Models
Explore more models from MiniMax
MiniMax-M2.5-Lightning
Ultra-fast variant of M2.5 generating 100 tokens per second at $1/hour continuous operation
MiniMax-M2.1
Lightweight coding-focused model with strong multi-language programming and agentic capabilities
MiniMax-M2
MiniMax's general-purpose foundation model with strong tool use and deep search capabilities
