Xiaomi just took the open model crown. mimo-v2.6, out september 22, scored 46 on artificial…
xiaomi just took the open model crown. mimo-v2.6, out september 22, scored 46 on artificial analysis' index, ahead of kimi k3 and qwen 3.8 max, after 30 steps of reinforcement learning.
the open leaderboard has a phone maker on top.
Context
Xiaomi's MiMo announcement for V2.6 says MiMo-V2.6-Pro scores 46 points on the Artificial Analysis Intelligence Index, surpassing Kimi K3 and Qwen3.8 Max. Artificial Analysis's own page for the model independently lists an Intelligence Index of 46, open weights and an MIT licence, and its comparison page shows 46 against 45 for the Qwen model compared. The Hugging Face model card describes a sparse mixture-of-experts model with 1.02T total and 42B active parameters, a 1M-token context and GRPO reinforcement learning. Xiaomi says that in under 6 days the Flash and Pro models each completed 30 reinforcement-learning steps.
The 46 is a third-party measured score on that one index. Ahead of Kimi K3 and Qwen 3.8 Max is Xiaomi's phrase, the margin against the Qwen model read is one point under that site's configuration, and the Kimi K3 variant seen in a snippet was Kimi K3 (Low), so which Kimi K3 setting is meant is unverified. Crown is bounded to that one index, and other leaderboards were not read. The 30 steps are vendor-reported. Xiaomi's release and the press are dated September 22 while Artificial Analysis says September 21, and the MIT licence comes from Artificial Analysis since the model card's licence field was not read. The open leaderboard has a phone maker on top is the author's take.
Watch next
- Independent runs and other leaderboards.
Sources
- Xiaomi MiMo: V2.6 announcementmimo.mi.com
- Artificial Analysis: MiMo-V2.6-Proartificialanalysis.ai
- Hugging Face: MiMo-V2.6-Pro-RLhuggingface.co
- Artificial Analysis: MiMo-V2.6-Pro vs Qwen3.8 Maxartificialanalysis.ai
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 22 September 2026 at 20:39 IST. Sources are the papers and datasets the note draws on.
View the original post ↗Embed this note
More notes
The air is now being asked to keep its own ledger
the air is now being asked to keep its own ledger: ecmwf’s aifs compo becomes the first ai model to forecast atmospheric composition globally every three hours, cleanair simulates 365 days of pm2.5 over china in ten seconds, and a unified framework maps six pollutants at one kilometer across the whole country. the air now files its own composition report.
read the note →The current is now being asked to draw its own map
the current is now being asked to draw its own map: china’s langya 2.0 predicts six ocean phenomena including internal waves and mesoscale eddies, a deep net called wenhai resolves eddies globally with air sea flux formulas built in, and scripps infers surface currents from the way temperature patterns deform in satellite images. the ocean now files its own circulation report.
read the note →The soil is now being asked to report its own carbon
the soil is now being asked to report its own carbon: a nix color sensor paired with generative data augmentation predicts soil organic carbon without a lab, random forest drives 74 percent of soil health mapping studies, and sentinel 2 tracks five year carbon change across france and italy from 922 samples. the dirt now files its own carbon account.
read the note →