LiveCodeBench Pro
2 models tested · Updated 2026-02-19 · Verified sources only
Gemini 3.1 Pro leads at 2887.0Elo
1
Google · Official/Google DeepMind · 2026-02-19
LiveCodeBench Pro uses competitive programming problems from Codeforces, ICPC, IOI. Vastly outperforms GPT-5.2 (2393) and Gemini 3 Pro (2439).
2887.0Elo
2
Meta · Blog/Meta · 2026-04-08
Competitive coding. Sits just behind GPT-5.4 and Gemini 3.1 Pro on Artificial Analysis Coding Index.
80.0%