🤝 Open to Collab

Mike Ravkine PRO

mike-ravkine

49 27 694

the-crypt-keeper

AI & ML interests

LLM Research / Development / Evaluation

Recent Activity

posted an update 3 days ago

https://huggingface.co/prism-ml/Ternary-Bonsai-2-27B-gguf at PQ2 is remarkable - an even more incredible feat of engineering then it's predecessor. I was impressed as hell with v1 but it suffered some truncation issues - they are essentially gone here. The entire task profile in both accuracy and token utilization has shifted *much* closer to the Qwen3.8-27B BF16 baseline. 👏 On the r12 👑 Leaderboard, Bonsai-2 lands at #24 significantly improving over the v1 #58 and landing ahead of many models several orders of magnitude larger in both reasoning performance and efficiency.

reacted to Datdanboi25's post with 🔥 4 days ago

THE SLM FRONTIER ADVANCES! https://huggingface.co/bench-labs/cagliostro-v3 just hit an Intelligence Index of 26.13 on the https://huggingface.co/spaces/AxiomicLabs/Open_SLM_Leaderboard a 146M-param model trained completely from scratch on a single consumer GPU. That's 2nd place overall, and as far as I can tell, the most capable SLM trained on consumer hardware to date. Beating SmolLM-135m on 1/8th of the data is just silly levels of efficiency. Big congrats to the @BenchLabs team and specifically @TobiasLogic !

liked a model 6 days ago

kacaforyah/Opti-27B

View all activity

Organizations

Posts 22

Post

prism-ml/Ternary-Bonsai-2-27B-gguf at PQ2 is remarkable - an even more incredible feat of engineering then it's predecessor.

I was impressed as hell with v1 but it suffered some truncation issues - they are essentially gone here. The entire task profile in both accuracy and token utilization has shifted *much* closer to the Qwen3.8-27B BF16 baseline. 👏

On the r12 👑 Leaderboard, Bonsai-2 lands at #24 significantly improving over the v1 #58 and landing ahead of many models several orders of magnitude larger in both reasoning performance and efficiency.