HumanEval.org

Benchmarks

LiveBench — Reasoning

last retrieved 2026-09-08 21:40 UTC

Third-party scores shown as context with provenance — our own arena ratings from blind human votes stay alongside every row and remain the primary signal.

ModelLiveBench — ReasoningObservedSource
GPT-6 AstraOpenAI92.72026-09-07source ↗
Claude Fable 5.1Anthropic91.72026-09-07source ↗
GPT-5.6 SolOpenAI91.72026-09-07source ↗
Claude Opus 5Anthropic91.22026-09-07source ↗
Kimi K3Moonshot AI90.72026-09-07source ↗
GPT-5.6 TerraOpenAI90.62026-09-07source ↗
Grok 4.6SpaceXAI90.52026-09-07source ↗
Claude Fable 5Anthropic89.72026-09-07source ↗
Gemini 3.8 FlashGoogle89.32026-09-07source ↗
Claude Opus 4.8Anthropic89.22026-09-07source ↗
Claude Sonnet 5Anthropic88.72026-09-07source ↗
Qwen3.8 Max (0902)Alibaba Qwen88.22026-09-07source ↗
GPT-5.4OpenAI88.12026-09-07source ↗
Claude Opus 4.7Anthropic87.22026-09-07source ↗
DeepSeek V4 Pro (0813)DeepSeek85.82026-09-07source ↗
GLM-5.3Z.ai85.82026-09-07source ↗
GPT-5.6 LunaOpenAI85.62026-09-07source ↗
Gemini 3.1 Pro PreviewGoogle84.02026-09-07source ↗

— = not yet evaluated by HumanEval (no eligible arena votes). External values are the newest observation per model; hover a value for source notes and a rating for its 95% CI. Unit: score, higher is better.

Not enough overlap with our arena ratings for a correlation scatter yet — fewer than two scored models are rated.