PrismML hopes its tiny LLM will change how we all use AI
PrismML released Bonsai 2 27B, compressing Alibaba's Qwen 27B to 5.9 GB with ternary weights while retaining 98% of benchmark performance.
PrismML, a Caltech-founded startup with a $22.25 million seed round from Khosla Ventures, Cerberus Capital, and Caltech, released Bonsai 2 27B, which compresses Qwen3.8 27B down to 5.9 GB, a 9-10x memory reduction small enough for a PC or high-end smartphone. The model uses ternary weights (+1, -1, 0) instead of 16-bit values and matches 98% of Qwen's aggregate benchmark scores, up from 95% for the original Bonsai, which has been downloaded over 11 million times. CEO Babak Hassibi said upcoming releases will target compressed models in the several-hundred-billion-parameter range, and the startup is rumored to be in talks with Apple.
- Bonsai 2 27B compresses Qwen3.8 27B to 5.9 GB, a 9-10x reduction
- Ternary weights (+1, -1, 0) replace standard 16-bit weight values
- Retains 98% of Qwen's aggregate benchmark scores vs 95% for Bonsai 1
- Backed by $22.25M seed from Khosla Ventures, Cerberus Capital, Caltech
- Next releases target several-hundred-billion-parameter compressed models
- Vendors
- PrismML
- Products
- Bonsai 2 27BQwen3.8 27B
- AI models
- Bonsai 2 27BQwen3.8 27B
From the original article
3 min read at techcrunch.com
If AI lab PrismML isn’t on your radar yet, it should be — not because it’s raised gobs of money (it hasn’t yet, just a $22. 25 million seed round), but because of the technical minds involved and the potentially industry-changing tech it’s developing. PrismML is betting that capable, high-performing, reasoning large language models don’t, in fact, have to be large.
By Julie Bort · techcrunch.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.