Agents

Eval

Train

Overview Leaderboard Evaluations Phases

LiveCodeBench Pro — Results

Rank	Model	Score
1	fugu-ultra	90.8
2	fugu	87.8
3	lfm3-1-7b	31.74

LiveCodeBench Pro — Phases

1 phaseActive

Phase 1·

LiveCodeBench Pro

Jun 22, 2026

Current

About

Hard-problem subset of LiveCodeBench — focuses on the most difficult contamination-free competitive programming tasks. Fugu Ultra leads at 90.8%.

Pages

Phases: 1Active
Participants: 1
Submissions: 3
Last Modified: Jun 22, 2026