Xiaomi · MiMo-V2.6
MiMo-V2.6-Flash
Smaller open-weight (MIT) 309B/15B-active omni-modal MoE from Xiaomi with a 1M context, positioned as a low-cost agent model.
- AA Intelligence
- —
- Input / output per 1M
- $0.14/ $0.28
- Output speed
- —
- Context window
- 1.05M tokens
Coding
- Terminal-Bench 4.0#16 of 8828.8%
Leader: Claude Mythos 5.1 at 60.9%
- Terminal-Bench 2.x#11 of 10387.6%
Leader: Claude Fable 5.1 at 91.4%
Agents and tool use
- OSWorld-Verified#6 of 1680.8%
Leader: Kimi K3 at 84.8%
Other published results
| Result | Score | Reported by | Source |
|---|---|---|---|
| DeepSWE v1.1 | 67.9% | vendor | Hugging Face: XiaomiMiMo/MiMo-V2.6-Flash-RL ↗ |
| ProgramBench | 26% | vendor | Hugging Face: XiaomiMiMo/MiMo-V2.6-Flash-RL ↗ |
| MiMo Code Bench | 61.2% | vendor | Hugging Face: XiaomiMiMo/MiMo-V2.6-Flash-RL ↗ |
| AutomationBench v1.0.6 | 52.3% | vendor | Hugging Face: XiaomiMiMo/MiMo-V2.6-Flash-RL ↗ |
| Toolathlon-Verified | 73.6% | vendor | Hugging Face: XiaomiMiMo/MiMo-V2.6-Flash-RL ↗ |
| Agents' Last Exam | 27.6% | vendor | Hugging Face: XiaomiMiMo/MiMo-V2.6-Flash-RL ↗ |
| JobBench | 61.2% | vendor | Hugging Face: XiaomiMiMo/MiMo-V2.6-Flash-RL ↗ |
| CyberGym | 95.1% | vendor | Hugging Face: XiaomiMiMo/MiMo-V2.6-Flash-RL ↗ |
| MiMo Cyber Bench | 77.2% | vendor | Hugging Face: XiaomiMiMo/MiMo-V2.6-Flash-RL ↗ |
| ExploitGym | 6% | vendor | Hugging Face: XiaomiMiMo/MiMo-V2.6-Flash-RL ↗ |
| ExploitBench | 25.3% | vendor | Hugging Face: XiaomiMiMo/MiMo-V2.6-Flash-RL ↗ |
| SEC Bench Pro | 47.5% | vendor | Hugging Face: XiaomiMiMo/MiMo-V2.6-Flash-RL ↗ |
| MiMo VisualCoding | 71.5% | vendor | Hugging Face: XiaomiMiMo/MiMo-V2.6-Flash-RL ↗ |