ByteDance Releases Math Reasoning Model Seed Prover 1.5, Reaches IMO Gold Medal Threshold
Want to read in a language you're more familiar with?
ByteDance's Seed Prover 1.5 model generated IMO 2025 proof code scoring 35/42 points, meeting the gold medal threshold. The team will open API access to researchers for further exploration in formal mathematics and AI.
ByteDance’s Seed research team has officially released Seed Prover 1.5, a model purpose-built for formal mathematical reasoning. The new version marks a major leap in Olympiad-level problem solving.
According to the team, Seed Prover 1.5 generated complete, compilable Lean proof code for the first five problems of the 2025 International Mathematical Olympiad (IMO) within 16.5 hours. When converted to the IMO scoring system, the model achieved 35 out of 42 points, meeting the historical gold medal cutoff.
Compared with its predecessor, Seed Prover 1.5 shows a substantial improvement in handling complex formal proofs. The team has publicly released the technical report and announced plans to open API access, inviting researchers in mathematics and AI to test the model and further advance the intersection of formal mathematics and artificial intelligence.
Source : AI daily
