MiMo-V2.6-Flash

LLM Stats ranks MiMo-V2.6-Flash at position 31 with a composite score of 45.7 at a blended price of $0.15/M. Its capability tier breakdown shows Top 10% in Coding (22 of 273), Average in Tool Calling (29 of 200), and Below top half in Chat (115 of 121). Quality tracker data begins September 22, 2026, consistent with the model's release window. Benchmark data from Xiaomi's official scorecard shows MiMo-V2.6-Flash at 0.95/1 on CyberGym (rank 1), 0.88/1 on Terminal-Bench 2.1 (rank 9), and scored on OSWorld-Verified. The LLM Stats page tracks the model's performance across conversation depth and cost efficiency, placing it between DeepSeek-V4-Flash and DeepSeek-V4.1-Flash on price-per-score.

According to Command Code's catalog page for MiMo V2.6 Flash (model id xiaomi/mimo-v2.6-flash), the model was released September 21, 2026, with a 1.05M token context window and native multimodal support. Its Artificial Analysis Intelligence Index is 37.9, with sub-scores of 74.3 for long-context reasoning and 51.3 for SciCode (scientific coding). API rates are $0.14/M input, $0.28/M output, and $0.0028/M cache read. The page also reports a blended price of $0.18/M and agent-loop cost of $0.04/M for MiMo V2.6 Flash, positioning it as the cheaper stablemate to MiMo V2.6 Pro ($0.43/$0.87 per M). Coding Index is not yet scored for this model. The catalog ranks it at 37.9 intelligence against 70 scored models.

On September 22, 2026, Xiaomi released and open-sourced the MiMo-V2.6 series, which includes MiMo-V2.6-Flash as the efficient, balanced variant alongside MiMo-V2.6-Pro. According to Xiaomi's announcement, MiMo-V2.6-Flash strikes the best balance between intelligence, efficiency, and cost. The series is described as natively omnimodal, and the page provides Xiaomi-reported benchmark scores including 67.9 on DeepSWE v1.1 and 71.5 on MiMo Visual Coding in-house. Xiaomi also reported that MiMo-V2.6-Flash achieved 95.1 on CyberGym (leading reported models), 73.6 on Toolathlon-verified, and 26.0 on ProgramBench. The MiMo-V2.6-Pro-UltraSpeed variant is rolled out separately for up to 20x faster output speed at the same quality. The release marks Xiaomi's continued exploration of scaling RL compute on verifiable, complex tasks for the MiMo family.