DigestAI news desk

Cut through the AI noise.

Listen
Generative AI & Models4 min read

IQuestLab releases IQuest-Q1, a 320B-parameter MoE model for coding agents

IQuestLab announced IQuest-Q1, a mixture-of-experts model with about 320 billion parameters, designed for agentic coding, reasoning, and multi-step tool use. The model activates roughly 15 billion parameters per token and lacks native multimodal capabilities. It is available via an OpenAI-compatible API, with deployment instructions provided for SGLang and vLLM frameworks.

1 source primary source

Key points

  • IQuest-Q1 is a 320B-parameter MoE model from IQuestLab for coding agents, with 15B active per token
  • OpenAI-compatible API deployment supported via SGLang or vLLM, with 512K-token context limit
  • Model lacks multimodal input and requires manual verification of generated code and reasoning

IQuest-Q1 scored on benchmarks like mini-SWE-agent and IQuest-CLIBench, though no external validation has been confirmed. The lab notes it remains an early-stage model with limitations in output reliability, tool integration, and real-world CLI task handling. Users are advised to review generated code and verify results manually. The model’s context window is capped at 512K tokens, and the lab recommends specific settings for optimal performance, including temperature 1.0, top-p 0.95, and top-k 20.

Model page: IQuest-Q1 →

Full story from huggingface.co · via Reddit AI communities primary sourceOpen source ↗

IQuestLab/IQuest-Q1 · Hugging Face

huggingface.co · 29 September 2026

Loading the full article…

This text was published by huggingface.co. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗

Coverage and discussion

1source
Topics · follow one to build your own front page
IQuestLabIQuest-Q1Claude Code 2.1.140Codex 0.142

The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.

Comments

via GitHub Discussions

More in Generative AI & Models

All →

Related stories

hot Meta launches enterprise AI platform, hires MongoDB CEO Chirantan Desai to lead it new Anthropic reportedly values itself at $2 trillion in leaked IPO prospectus hot OpenAI's GPT-6 Astra: Changing Work and Research new EliseAI raises $350M at $4B valuation to expand AI in housing and healthcare hot NVIDIA launches Open Agent Safety Platform with OpenShell and Sentry to isolate AI agents in milliseconds new NYC Council subpoenas SpaceX over AI risks for October hearing hot AMD agrees to acquire Fei-Fei Li's World Labs for $8.2B in all-stock deal new OpenAI scraps GPT-6.1 Astra release over safety concerns