@NielsRogge: That's right, a 3.9GB file getting 99.2% on Math 500 A dataset once introduced by @OpenAI in the paper "Let's verify st…
Summary
PrismML's Bonsai-27B model achieves 99.2% on the Math 500 dataset, outperforming DeepSeek-R1 and Kimi K2.
View Cached Full Text
Cached at: 07/16/26, 04:05 AM
That’s right, a 3.9GB file getting 99.2% on Math 500 🤯
A dataset once introduced by @OpenAI in the paper “Let’s verify step by step”.
It beats models like DeepSeek-R1 and Kimi K2 on it https://t.co/QZ5vt74LQ6
Niels Rogge (@NielsRogge): The technical report of @PrismML’s Bonsai-27B is top-trending on Papers with Code.
Find it here, alongside evals, HF models and more:
Similar Articles
prism-ml/Bonsai-27B-gguf
Prism ML releases Bonsai-27B-gguf, a 27-billion parameter language model with binary (1.125-bit) weights, achieving a ~14x size reduction while retaining ~90% of FP16 reasoning performance. It runs on consumer hardware with high throughput.
@TheAhmadOsman: Been playing with @PrismML's new model that turned Qwen 3.5 27B into a sub-4GB and sub-6GB weights and I am impressed C…
PrismML released a compressed version of Qwen 3.5 27B that fits in sub-4GB and sub-6GB memory, enabling impressive local AI performance.
I benchmarked PrismML's 1-bit Bonsai-8B against IBM's Granite on CPU tool calling. The 1-bit model won, but only with grammar-constrained decoding
An independent benchmark of PrismML's 1-bit Bonsai-8B against IBM's Granite and other models on CPU tool calling shows that with grammar-constrained decoding, Bonsai-8B achieves a 92% pass rate, outperforming larger models, but fails without constraints. Granite is the best raw model at 72%.
@PrismML: Today, we’re announcing Bonsai 27B: the first 27B-class model to run on a phone. Bonsai 27B is the new multimodal flags…
PrismML announces Bonsai 27B, the first 27-billion parameter model able to run on a phone, with ternary and 1-bit variants (5.9 GB and 3.9 GB respectively) that enable multi-step reasoning and agentic workflows on local devices, all open-sourced under Apache 2.0.
@populartourist: I can run a 27B instead of a 9B on my budget laptop. That's mind blowing.
PrismML announces Bonsai 27B, a multimodal model based on Qwen3.6 27B that can run on a phone, enabling local multi-step reasoning, tool use, and long-context workflows.