IQuest Research has released IQuest-Q1, a 320B parameter model with a sparse MoE architecture that activates 15B parameters. This model demonstrates impressive capabilities, including generating a functional HTML game from a simple prompt and identifying a subtle bug in RL training data that caused issues with text decoding and trajectory stitching. IQuest-Q1 has also shown strong performance on various benchmarks such as NL2Repo, CyberGym, Terminal-Bench 2.1, DeepSWE v1.1, and JobBench, positioning it as a capable model within its parameter class. AI
IMPACT Demonstrates advanced capabilities in creative generation and complex debugging, potentially accelerating AI development workflows.
RANK_REASON New model release from a research lab with detailed capabilities and benchmark performance. [lever_c_demoted from frontier_release: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
