Xiaomi MiMo-V2.5 is an open-source Xiaomi model for agentic coding, long-horizon tasks, and 1M-token context with MoE and hybrid attention.
Capabilities, design details, and architectural traits
MiMo-V2.5 is built around long, structured work. It is presented as an open-source model for agentic coding and other long-horizon tasks that need sustained context.
| Differentiator | Documented trait |
|---|---|
| 1M-token context | Supports up to 1M tokens of context. |
| Mixture-of-Experts design | Uses a 1.02T-parameter MoE model with 42B active parameters. |
| Hybrid attention + MTP | Uses hybrid attention and Multi-Token Prediction from the MiMo-V2-Flash design. |
| Open-source release | The MiMo-V2.5 series is open-sourced under the MIT license. |
The model is aimed at tasks that stay active across long spans of text and tool use. The documented focus is on agentic coding and long-horizon execution rather than a short chat exchange.
MiMo-V2.5 belongs to a family that includes Pro, Omni, Flash, and TTS variants. The series is routed through Xiaomi MiMo with unified model naming across products.
Independent evaluations · Artificial Analysis
Evaluate specifications, pricing, and independent benchmark indices
| Model Details | |||
|---|---|---|---|
| General Info | |||
| Provider | Xiaomi | Anthropic | Anthropic |
| Release Date | April 22, 2026 | July 24, 2026 | June 9, 2026 |
| Knowledge Cutoff | - | May 2026 | - |
| Context & Limits | |||
| Context Window | 1.0M Best Context Window | 1M | 1M |
| Pricing (per 1M tokens) | |||
| Input Pricing | $0.14 Best Input Pricing | $5 | $10 |
| Output Pricing | $0.28 Best Output Pricing | $25 | $50 |
| Modalities | |||
| Inputs | textaudioimagevideo | textimage | textimagefile |
| Outputs | text | text | text |
| Benchmarks (0-100) | |||
| Intelligence Index | 38.0 | 63.1 Best Intelligence Index | 62.1 |
| Coding Index | 56.8 | 78.0 Best Coding Index | 76.5 |
| Agentic Index | 24.4 | 59.2 Best Agentic Index | 56.6 |
Graduate-level reasoning and expert Q&A evaluation.
Extremely difficult logical reasoning and knowledge.
Logical reasoning over long context windows.
Independent evaluation data provided by Artificial Analysis. To view the latest benchmarks and full details, visit their official site.
