The reasoning race is moving to the edge. meta shipped mobilellm-r1 on hugging face, sub-1b models…
the reasoning race is moving to the edge. meta shipped mobilellm-r1 on hugging face, sub-1b models from 140m to 950m params tuned for math and coding reasoning on-device.
the interesting benchmark gap is now what a phone can run, not what a cluster can.
Context
Meta's MobileLLM-R1 repository says the models were released on Hugging Face on 12 September 2025, in sizes from 140M to 950M parameters, with the 950M pre-trained on about 2T tokens. Meta reports it as comparable to Qwen3 0.6B on MATH, GSM8K, MMLU and LiveCodeBench. The repo was created 23 September 2025, the technical report is dated 29 September 2025, and the repo notes ICLR 2026 acceptance. The license is listed as Other.
The release is about a year before the note's date, and no September 2026 Meta release of this model was found, so the note reads as current news when the evidence is from 2025. The benchmark comparison is Meta's own. No phone benchmark was in the text read, so what a phone can run is the author's thesis and not a measured claim.
Watch next
- A 2026 successor in Meta's repo and independent on-device latency tests.
Sources
- facebookresearch/MobileLLM-R1 (GitHub)github.com
- MobileLLM-R1-950M (Hugging Face)huggingface.co
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 20 September 2026 at 05:17 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 →