One of my favorite small local coding models got faster.
This is a dev post classified by AGK Intelligence as AI dev tools (a launch), kept by the Dev Radar because it carries real work, not commentary.
One of my favorite small local coding models got faster. JetBrains released a new MTP head for Mellum2.1 Thinking. And this model already surprised me because it beat Qwen in my testing. Mellum2.1 ... 🧠 12B total / 2.5B active 📚 131K context 💾 Q4_K_M + MTP → ~8.3GB 🔓 Apache 2.0 💻 coding + tools + repo agents Take note that JetBrains didn't replace the model. The underlying Mellum2.1 weights are the same. They added a small MTP head that tries to predict several future tokens ahead. The main model verifies them. JetBrains says ⚡ 1.4–1.8× faster decoding 🎯 ~90% draft-token acceptance
Posted by David Hendrickson (12.6k followers) 1 h ago · 5 likes · 351 views · view the original post on X. Kept by the Dev Radar as AI dev tools.
More dev work like this
- I made a UI library that looks like it was drawn by hand. — @kevin_t_ngo
- Find AI tools without opening another dozen tabs. — @DanKornas
- OpenAI quietly shipped Python SDK v3.42.0. — @akshay81844
- Grokbot 的替代品来了，这玩意儿今天刚上了 github trending，截止到现在已经 100K star 了 — @off_thetarget
- Qwen has landed in TensorFold Living Weights. — @ashxhart
- Give your local models a chat interface you can use on your phone. — @DanKornas
- 同时让几个 AI 干活，窗口开了一排，哪个在写代码、哪个等着补材料，来回切几次就容易弄混。 — @GitHub_Daily
- Quantum Generative Adversarial Networks! #BigData #Analytics #DataScience #AI… — @gp_pulipaka
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. 31.2k 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 04:11 UTC. Full method.