Mimo-V2.6-Flash
AI overview
Sign in and the AI will write an overview from our coverage.
Headlines · 4
- Arena revises Alignment Index methodology and model rankings
Arena updated the Alignment Index methodology, saying its revised failure signals provide broader coverage and more reliable judgments. Under the new rankings, OpenAI models hold all seven top positions, Claude Fable 5.2 debuts eighth with a score of 78.2, and Claude Opus 5.5 falls from sixth to seventeenth. Kimi K3.1 Pro ranks ninth, while most open-source models trail closed releases.
arena (X) · 🔥 3 - Arena finds cheaper coding models are rapidly catching up
Arena analyzed 105 models from 24 labs across 44 weekly Code Arena: WebDev leaderboards, covering December 2025 through October 23, 2026. It found that new model additions rose from about five per month last winter to 12–14 per month from July through October, while top scores improved three times faster from May to October. Xiaomi’s MiMo-V2.6-Flash outscored Claude Opus 4.7 at less than one-twentieth the price, and October’s best model had an estimated 65–80% win rate against April’s best at matching answer prices.
arena (X) · 🔥 0 - Xiaomi takes MiMo-V2.5-Pro and V2.5 offline, routing API calls to V2.6
Xiaomi's MiMo documentation states that mimo-v2.5-pro and mimo-v2.5 were officially taken offline at 10:00 UTC+8 on October 21, 2026, and that requests have been automatically routed to mimo-v2.6-pro and mimo-v2.6-flash since 18:00 UTC+8 on October 14, IT之家 reported. V2.5-Pro accepted only text input, while its replacement V2.6-Pro supports multimodal understanding, and the replacement table says API parameters are fully compatible. The documentation also notes that in thinking mode the V2.6 models do not accept custom temperature or top_p (fixed at 1.0 and 0.95), that V2.6-Flash's default max_completion_tokens is 131,072 versus 32,768 for V2.5, and that MiMo-V2.6 keeps V2.5 API pricing.
IT之家 AI · 🔥 0 - Xiaomi's MiMo-V2.6 report details $2.6M single-run RL post-training
Xiaomi's MiMo team published a technical report, "MiMo-V2.6: Scaling Reinforcement Learning Towards Self-Improvement," covering two sparse MoE models: MiMo-V2.6-Pro (1.02T total parameters, 42B activated) and MiMo-V2.6-Flash (310B total, 15B activated). The report says each RL step samples 1,568 prompts and roughly 25,000 agent trajectories totalling 2.7–3.7B tokens, with RL post-training costs of about $2.6M for Pro and $900K for Flash. The team also open-sourced more than 7,000 RL environments plus an end-to-end RL framework.
InfoQ 中文 AI&大模型 · 🔥 0
Experience and discussion from the community
Share my Mimo-V2.6-Flash experienceAsk about Mimo-V2.6-Flash
Nobody has shared their experience with Mimo-V2.6-Flash yet.