MiMo-V2.6-Flash

Xiaomi's MiMo team released MiMo V2.6 on September 22, 2026 as a family of open-weight, natively omnimodal language models that accept text, images, video, and audio in a single checkpoint, with every variant shipping a 1 million-token context window. The series is framed around the tagline "Scaling Reinforcement Learn Within that family, MiMo-V2.6-Flash is positioned as the smaller, substantially cheaper sibling designed for high-volume production workloads while retaining the 1-million-token context and native multimodal capabilities shared with Pro. Both variants are MIT-licensed open-weight releases that can be downloaded from Hu

Xiaomi released the MiMo-V2.6 family on September 21, 2026, with MiMo-V2.6-Pro debuting as the top-performing open-weight model on Artificial Analysis' Intelligence Index at a score of 46, ahead of xAI's Grok 4.6 (44) and Google's Gemini 3.8 Flash (41), and tied with Grok 4.7 which also scored 46 the same day. Pro is M Alongside Pro, Xiaomi launched MiMo-V2.6-Flash, a smaller and substantially cheaper variant priced at $0.14 per million input tokens and $0.28 per million output tokens. The Flash model retains the same 1-million-token context window and native multimodal (text, image, audio, video) capabilities as Pro while targeting

LLM-Stats' MiMo-V2.6-Flash profile places the model at composite rank 31 on the LLM Stats Score, with capability tiers of Coding 22 of 273 (top 10%), Tool Calling 29 of 200, Reasoning 45 of 369, Vision 47 of 213, and Chat 114 of 120 (below top half). Blended price per million tokens is listed at $0.15, and per-dataset The Quality Tracker shows only a single community vote with a 0.00σ change over 7 and 30-day windows, indicating very thin independent corroboration so far for the Flash variant's performance claims. Pricing figures, benchmark rows, and tier standings on this aggregator page are explicitly scoped to MiMo-V2.6-Flash rat

BenchLM's dedicated MiMo-V2.6-Flash profile lists Xiaomi's API pricing at $0.14 per million input tokens, $0.28 per million output tokens, $0.003 per million cached input tokens, and a blended figure of $0.21 per million tokens, with a 1-million-token maximum context window. The page reports 14 published benchmark rows Several benchmark categories on the BenchLM page carry "Not measured" or "Not eligible" labels, including Reasoning, Multimodal, Knowledge, Multilingual, Instruction Following, and Math, meaning BenchLM does not assign MiMo-V2.6-Flash a public comparative rank despite the pricing and context evidence. The data is times