I've been building a custom autonomous coding agent loop powered by GLM-5.1, and I just ran it against the LiveCodeBench Lite dataset (Python split).

Here is the breakdown over a blind 369-problem run:

Total: 348/369 passed (94.3%) Easy: 138/141 passed (97.9%) Medium: 152/156 passed (97.4%) Hard: 58/72 passed (80.6%)

I'm not just building this for the leaderboard. I am launching this to drive immediate humanitarian impact. 100% of the profits from 11 months of the year from this agent will be donated directly to Palestinian refugees.

I'm currently looking for fast angel funding, compute sponsorships, or strategic partners to help scale this ASAP.

You can find the raw JSONL with code outputs here: https://github.com/UnivenceAI/Univence-benchmarks/tree/main/Z%20AI/GLM-5.1

If you want to try it out you can at https://univence.com

You can also follow our progress on X at https://x.com/UnivenceAI

If you want to contact us you can at univenceai@gmail.com which is our temporary email until we've setup a custom domain email address.

I would love any feedback from the community, and my DMs/inbox are open if you are interested in supporting the mission.