I just released an open-source, python-based autonomous theorem prover , Ensemble Prover, on GitHub. It combines LLM proof search with Lean verification.
To date, it has solved 65 PutnamBench problems during development.
LLMs tested via API: GPT-5.2,GPT-5.6 Luna-Pro, DeepSeek-V4-Flash, DeepSeek-V4-Pro, and Qwen3.7-Max.
Development is ongoing. Would love to see others use it to attempt unsolved problems in parallel.
I just released an open-source, python-based autonomous theorem prover , Ensemble Prover, on GitHub. It combines LLM proof search with Lean verification.
To date, it has solved 65 PutnamBench problems during development.
LLMs tested via API: GPT-5.2,GPT-5.6 Luna-Pro, DeepSeek-V4-Flash, DeepSeek-V4-Pro, and Qwen3.7-Max.
Development is ongoing. Would love to see others use it to attempt unsolved problems in parallel.