MiMo v2.5 Pro
XiaomiFlagship
ThinkingTool UseStructured Output
About this model
MiMo-V2.5-Pro is Xiaomi’s flagship model, delivering strong performance in general agentic capabilities, complex software engineering, and long-horizon tasks, with top rankings on benchmarks such as ClawEval, GDPVal, and SWE-bench Pro....
Performance Tier
Flagship
MiMo v2.5 Pro is a flagship model from Xiaomi : the most capable in their lineup.
Best-in-class model from this provider. Highest performance across benchmarks, ideal for demanding tasks.
Pricing
This model is included in Elosia plansAffordable
Low cost. Suitable for sustained use and high-volume interactions.
| Type | per 1M tokens |
|---|---|
| Input (prompt) | $0.435 |
| Output (completion) | $0.870 |
| Cache read | $0.0036 |
Capabilities
Context Length1.0M
Max Output Tokens131K
TokenizerOther
Inputtext
Outputtext
Release DateApril 22, 2026
Benchmarks
General Intelligence
MMLU
Not reported
GPQA Diamond
86.6%
Mathematics
MATH-500
Not reported
Programming
HumanEval
Not reported
SWE-bench Verified
78.9%
Reasoning
IFEval
Not reported
Humanity's Last Exam
33.8%
Agentic
SWE-bench Pro
57.2%
Recommended Use Cases
CodingAnalysisResearchMathematics
Strengths
- MoE architecture (1.02T total, 42B active) under a permissive MIT license — self-hostable at a frontier capability tier
- Long-horizon agentic coding (SWE-bench Verified 78.9), sustaining 1,000+ sequential tool calls over the 1M-token context
- Token-efficient — comparable quality with 40-60% fewer tokens than Claude Opus 4.6, Gemini 3.1 Pro and GPT-5.4 (vendor-reported)
- Competitive agentic coding among open-weight models (SWE-bench Pro 57.2)
- Aggressive pricing at $0.44/M input and $0.87/M output — far below comparable frontier models
Limitations
- Text-only — no image, audio or video input (the omni-modal capability lives in the separate MiMo-V2.5)
- Headline scores are self-reported; independent verification is limited, and an external harness measured Terminal-Bench well below the vendor figure
- Slower inference than peers (~45-52 tokens/s) despite the low price
Resources
This model may use your data for training
