Unsloth's 2-bit DeepSeek beat their own reference on a math test, by 2 answers out of 480.
This is a dev post classified by AGK Intelligence as AI dev tools (an opinion), kept by the Dev Radar because it carries real work, not commentary.
Unsloth's 2-bit DeepSeek beat their own reference on a math test, by 2 answers out of 480. AIME25: 97.50 vs 97.08. 30 problems, 16 samples each. I will not tell anyone their 2-bit is better because of 2 answers. Every 95% bar on their chart is wider than that gap. People argue Unsloth vs GSQ here. In Unsloth's run the gap at equal size is small. Around 232 GB Unsloth is ahead by 5 answers, around 250 GB GSQ by 1. In that run Unsloth leads HMMT and Aider at equal size. I like both teams. GSQ-RCO is in my Qwen Flash-Next repo, I ran MiniMax on Unsloth's 2-bit. Nobody can call a winner from
Posted by Alexey Fateev (9.6k followers) 2 h ago · 10 likes · 669 views · view the original post on X. Kept by the Dev Radar as AI dev tools. Tools mentioned: deepseek-v4.1-flash-1017-gguf, Introducing Unsloth Studio.
More dev work like this
- Qwen3.8-27B decodes code at 53.6 tok/s on one 16 GB RTX 5060 Ti and 154.5 tok/s on a… — @niklaslenz_ai
- DevX with NeoAstra is just phenomenal 🤩 using CodeAlta and GPT Luna 6 for a couple of… — @laurentkempe
- computer-repair-skill 这个开源项目很接地气，值得推荐。 — @bkdgiffug
- Your AI workflow doesn’t need five separate windows. — @DanKornas
- Claude Sonnet 5.5 vs GPT-6.1 Sol for computer use. — @coastyai
- TensorFold 1.2.0 decodes Qwen3.8-27B code at 268.6 tok/s on an RTX 5090, level with 1.1.2. — @niklaslenz_ai
- DeepSeek V4.1 Flash 1017 for 4x DGX Sparks got a lot of updates, with support for… — @MiaAI_lab
- Vibe coding works when your agent has the right tools. — @apify
Every post is read and classified by AGK Intelligence: what it is, which market it belongs to, and whether the link is a real tool. 31.3k posts from 5k X accounts over the last 21 days, 8.9k tools, 12 markets. New posts are collected several times an hour and the full ranking is recomputed every hour — last update 2026-11-01 11:47 UTC. Full method.