IQuest-Q1: a free open-weight 320B model that diagnoses training issues and fixes its own code
LearnWithBishal · x · 2026-09-29
IQuest-Q1 is now available for free as an open-weight model on GitHub and Hugging Face, along with a technical report and blog.
The 320B-parameter model is pitched around autonomous engineering capabilities:
- Diagnosing its own training issues
- Locating bugs in code
- Applying fixes
- Verifying the fixes actually worked
Links are distributed via a comment-to-DM mechanic on X.
Related event: IQuest Releases Open-Weight 320B MoE Model IQuest-Q1 for Agentic Coding(7 posts)→
More from coding & agent
- Muse Code 1.4.0 login broken on Windows and headless Linux; fix shipped in R4302.1 — PawelHuryn · 2026-09-29
- Open-source Skill Makes AI Agents Ship Obsidian Plugins Like Real Products — vista8 · 2026-09-29
- Open-source Skill for Obsidian plugin development, installable with one command — vista8 · 2026-09-29
- Muse Code 1.4.0 hits triple bug reports: broken login, resume kills workflows, 60s hangs — PawelHuryn · 2026-09-29
- Respan launches Guardrails, claims top average F1 across 8 safety categories, free with its gateway — ycombinator · 2026-09-29
- AsideAI cuts memory/CPU use 60-93% and tokens 7-14x in latest update — garrytan · 2026-09-29