Unverified50% confidenceFactExact time
The Bonsai 2 model released by Prism ML claims to retain 98% of the performance of a 27-billion-parameter model within less than 8 GB of VRAM.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/4/2026
First Seen
Valid until: 1/2/2027
Sources
Related Entities
Related Claims
UnverifiedPrismML's Ternary Bonsai 2 applies ternary quantization to the Qwen3.8 27B model, has received 3.87 million downloads, compresses the language model from 54GB to 5.9GB — a 9.3x reduction — and officially claims it retains 98.2% of the FP16 model's capability, with an average score of 84.78 on 14 reasoning benchmarks.78% similarUnverifiedPrismo claims its solution activates all 27 billion parameters simultaneously and compresses the original 54 GB QW3.6 model to less than 4 GB, a compression ratio of over 13x, with claimed no impact on performance.70% similarUnverifiedNewer large-parameter models such as Flux have higher VRAM requirements, and the 3060 12GB can run quantized versions but with compromised speed.68% similarPartially VerifiedThe QLoRA paper verified that NF4 quantization combined with LoRA fine-tuning can reduce memory requirements from 780GB to 48GB on a 65B parameter model while maintaining performance close to full-precision fine-tuning66% similarUnverifiedAccording to community discussions, the DeepSeek-V4.1-Flash model's 1 million token context window occupies only approximately 900 MB of VRAM.65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/973445API
curl https://kongchang.com/api/v1/knowledge/claims/973445MCP
get_claim(id=973445)