List your company

Formal Theorem Proving on Combibench

48Solve RateSeed-Prover 1.5
-1.795211.132424.0636.9876Dec 19, 2025Jan 9, 2026Jan 31, 2026Feb 22, 2026Mar 16, 2026Apr 7, 2026Apr 29, 2026

Evaluation results

MethodLinks
Seed-Prover 1.5Compute Budget=10 H20 days / problem2025.1248
Seed-Prover 1.0 (medium)Compute Budget=18 H20 days / problem2025.1239
HilbertBackbone=Gemini 3.1 Pro, System=Agentic System2026.042.05
Goedel-Prover-V2-8BType=Open-source LLM2026.041.96
HilbertBackbone=Gemini 2.5 Pro, System=Agentic System2026.041.89
Goedel-Prover-V2-32BType=Open-source LLM2026.041.31
DreamProverBackbone=Gemini 3.1 Pro, System=Lemma Learning System2026.040.93
DreamProverBackbone=Gemini 2.5 Pro, System=Lemma Learning System2026.040.75
Gemini 2.5 ProType=Proprietary LLM2026.040.64
Gemini 3.1 ProType=Proprietary LLM2026.040.59
HilbertBackbone=GPT-5.3-Codex, System=Agentic System2026.040.58
DeepSeek-Prover-V2-7BType=Open-source LLM2026.040.42
Claude 4.6 OpusType=Proprietary LLM2026.040.41
DreamProverBackbone=GPT-5.3-Codex, System=Lemma Learning System2026.040.32
GPT-5.3-CodexType=Proprietary LLM2026.040.12