Prismml shipped bonsai 2 27b on september 17 with every weight reduced to a value in the set minus…
prismml shipped bonsai 2 27b on september 17 with every weight reduced to a value in the set minus one, zero, or one, packing a 27 billion parameter model into 5.9 gigabytes at 1.76 effective bits per weight. the compressed version keeps 98 percent of the full precision score, up from 95 percent in the previous generation.
the gap between quantized and full models is closing faster than the gap between frontier labs.
Context
PrismML's post of 17 September 2026 says Bonsai 2 27B, a ternary model based on Qwen3.8 27B, is 5.9 GB and more than 9x smaller than the full-precision baseline. It says the original Ternary Bonsai 27B kept 95% of aggregate performance and the new one keeps over 98% (98.2%), with a score of 83.9 against 85.4 for Qwen3.8 27B on a 20-benchmark aggregate.
All figures are PrismML's own, vendor-reported and not independent, and the 98% is an aggregate retention figure, not a per-task result. A Hugging Face MLX 2-bit card lists 8.60 GB, which is a different packaging and was not merged. That the gap between quantized and full models is closing faster than the gap between frontier labs is the author's opinion; no source read tests it.
Related work
- Earlier note on the same release ↗Same PrismML release.
Watch next
- Independent evaluations on the same harness.
Sources
- Bonsai 2 27B (PrismML, 17 Sep 2026)prismml.com
- PrismML launches Bonsai 2 27B (17 Sep 2026)prismml.com
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 18 September 2026 at 22:34 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 →