MiMo V2.6 Pro

Felo's September 24, 2026 blog confirms Xiaomi's MiMo V2.6 Pro is now live on Felo OpenAPI at $0.43/$0.87 per million tokens using the same API key and base URL as other providers. The post notes Xiaomi shipped weights, a technical report, and the reinforcement learning training code alongside the model, and reports th On DeepSWE v1.1, an out-of-sample long-horizon software engineering benchmark, the Pro model moved from 48.8 to 65.7 during the open RL phase. Xiaomi also open-sourced more than 7,000 reinforcement learning task environments spanning software engineering, vulnerability reproduction, knowledge-intensive work, and web de

Forkast's September 22, 2026 coverage confirms Xiaomi's MiMo-V2.6-Pro release, scoring 46 on the Artificial Analysis Intelligence Index v4.3 and tying Grok 4.7 to set a new ceiling for open-weight models. The team, led by Luo Fuli (formerly of DeepSeek), released both Pro and Flash variants with open weights on Hugging Technical details describe a frozen-router MoE with 1.02 trillion total parameters and 42 billion active, hybrid attention mechanisms, a 5-layer MTP speculative decoder, and fully asynchronous GRPO training using 1,568 prompts across 16 rollouts per step. The model supports a 1-million-token context window with native

Flowtivity's September 22, 2026 analysis describes Xiaomi's MiMo-V2.6 series as the strongest open-weights model family available, with MiMo-V2.6-Pro scoring 46 on the Artificial Analysis Intelligence Index — one point behind OpenAI's GPT-5.6 Sol at 47 — at approximately $0.13 per task. The Pro variant uses 1.02 trilli The analysis positions MiMo-V2.6-Pro on the intelligence-versus-cost Pareto frontier, noting that at comparable intelligence the Pro model costs roughly one-twentieth to one-sixtieth of leading international models according to Xiaomi's launch claims. The MIT license on the open weights is highlighted as a practical en

Independent analysis dated September 22, 2026, confirms Xiaomi's MiMo-V2.6-Pro is the highest-scoring open-weights model on the Artificial Analysis Intelligence Index, scoring 46 — tied with Grok 4.7 — compared with Claude Opus 5's 51, at roughly $0.13 per index task versus Claude Opus 5's $5.86. The piece separates Xi Specifications confirmed across Xiaomi docs, OpenRouter, and Artificial Analysis include a frozen-router Mixture-of-Experts architecture with 1.02 trillion total parameters and 42 billion active, a 1,048,576-token context window with 131,072-token max output, and native multimodal inputs (text, image, speech, video) wi