Xiaomi MiMo-V2.6 tops open-source AI leaderboard at 1/20th the price of Grok 4.7
卡尔的AI沃茨 · wechat · 2026-09-23
Xiaomi released three MiMo-V2.6 variants (Pro, Flash, Pro-UltraSpeed) just five days after livestreaming their RL training. Pro scores 46.32 on Artificial Analysis—first among open-source and Chinese models, tying Grok 4.7 released the same day, at roughly 1/20th the price. UltraSpeed outputs up to 20x faster than Pro.
Pricing stays aggressive: Pro costs ¥0.025/M cached input and ¥6/M output, Flash ¥2/M, with batch inference at half price—a 20x-60x gap vs comparable overseas models. The base model is unchanged (1.02T total, 42B active); gains come almost entirely from post-training. Notable benchmark results: third among open models on Terminal-Bench 4.0 (behind GLM-5.3 and Qwen3.8Max), near the top on physics-reasoning CritPT, and DeepSWE v1.1 jumped from 58.4 to 72.57.
The author's hands-on tests—website replication from raw screen recording via native video input, pelican-on-a-bicycle SVG animation, multimodal PPT generation from PDFs/audio/images, Blender box-fold modeling, a 3D ninja game, and a real project combining Jev for live data re-crawling with MiMo for parsing—suggest it's ready to move from 'good enough' to a daily-driver agent model.
Related event: Grok 4.7 Meets Xiaomi MiMo-V2.6 in a Same-Night Price War(5 posts)→
More from Models
- Grok Bot 1.0 Announced: Shareable Bots With Upgraded Collaboration — SpaceXSI · 2026-11-12
- Polymarket odds put Chinese AI lab at #1 by year-end at just 6% — Polymarket · 2026-11-12
- FutureHouse's deslopper trains Kosmos to write indistinguishably from human experts — proteinrosh · 2026-11-12
- Developers say Gemini 4 Carbon is notably better at favicon design and micro-interactions — AI_Andrew · 2026-11-12
- Gemini Decision 1 tops knowledge and reasoning evals on calibration, at a steep cost — ArtificialAnlys · 2026-11-12
- Gemini Decision 1 costs $345 per 1M tasks, 3x pricier than leader Drex Max v1.1 — ArtificialAnlys · 2026-11-12