Alibaba shipped qwen3.8-flash-next with a 51b parameter component designed to live in system ram…
alibaba shipped qwen3.8-flash-next with a 51b parameter component designed to live in system ram instead of gpu memory. the interesting part is not the count. it is that the inference graph now treats cpu as a tier of vram.
the line between model and context just got blurry.
Context
The Hugging Face model card for Qwen3.8-Flash-Next lists 125B parameters with 6B activated, plus a 51B n-gram embedding and 4B MTP. The card says the n-gram embedding requires less computation and is more amenable to offloading than Mixture-of-Experts, and suits memory-constrained accelerators.
The card supports a 51B n-gram embedding component and the offloading language. It does not state that the component is designed to live in system RAM, so that wording is the author's reading. The card date seen in search metadata is 26 August 2026, before the note, so the model was not new on 5 September. Treating CPU as a tier of VRAM and the model and context line blurring are the author's opinion.
Related work
- Qwen3.8-Flash-Next repository (GitHub) ↗Matches the 125B and 51B description, seen as a snippet only.
- Running Qwen3.8-Flash-Next locally (MortalApps) ↗Secondary, dated 15 September 2026, after the note; snippet only.
Watch next
- Qwen documentation on where the n-gram embedding is meant to run.
Sources
- Qwen3.8-Flash-Next model card (Hugging Face)huggingface.co
- Qwen3.8-Flash-Next technical reportraw.githubusercontent.com
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 5 September 2026 at 07:03 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 →