Coding

LiveCodeBench: 2026 AI Leaderboard

Competitive programming problems published AFTER the model's training cutoff.

What it tests

LiveCodeBench collects problems from LeetCode, AtCoder, and Codeforces, explicitly including only problems published after a model's knowledge cutoff to prevent training contamination.

How it is scored

Pass@1 rate on held-out-by-date competitive-programming problems. Difficulty bucketed Easy/Medium/Hard. Because the problem set rolls forward with time, scores are comparable year over year.

Why it matters

LiveCodeBench is the best available contamination-resistant coding benchmark. It discriminates between models that have memorized common algorithms and models that can actually solve novel problems under competitive constraints.

Leaderboard (1 models)

Sorted by LiveCodeBenchscore. Tier column shows the tool's overall AIToolTier rank, which blends this benchmark with pricing, features, and real-world usability.

#ModelTierLiveCodeBench score
1Kimi K2.5 (Moonshot)
Kimi K2.5 (1T/32B active MoE)
A74.1%

About LiveCodeBench

Creator
LMSYS & Berkeley, 2024
Unit
% (max 100)

Other benchmarks