LiveBench

LiveBench was built to answer a specific worry about every other benchmark on this list: that a widely used test eventually leaks into a model’s training data, so a high score reflects memorisation rather than ability. A team spanning Abacus.AI, NYU and the University of Maryland — including Tom Goldstein and Yann LeCun among its authors — published it in June 2024 with a fix: draw questions from sources too recent to have been trained on, such as that month’s arXiv papers, news articles and maths competitions, refresh the set on a rolling schedule, and grade every answer against an objective ground truth rather than a human or LLM judge’s opinion.

That design also sidesteps a second problem, the one that dogs arenas and LLM-judged benchmarks alike: a judge model can be gamed by style rather than substance. LiveBench’s 23 tasks, spread across reasoning, coding, agentic coding, mathematics, data analysis, language and instruction following, are checked against fixed answers instead. The tradeoff is that the whole question set is periodically replaced — the site had moved from its original 2024-06-24 release to a 2026-06-25 edition by the time of writing — which keeps scores comparison-resistant across versions in the way a static benchmark’s are not.

Unlike single-lab benchmarks, LiveBench is genuinely competitive across the field: every major frontier lab appears on its board, and the overall lead has passed back and forth between Anthropic (Claude 3.5 Sonnet at launch), Google (Gemini 2.5 Pro), OpenAI (o3, later GPT-5.5) and back to Anthropic — Claude Fable 5 leads the current 2026-06-25 set at 83.0, narrowly ahead of OpenAI’s GPT-5.6 Sol at 81.0, with strong open-weight entries from Moonshot, Alibaba and Abacus just behind. One subtlety is essential to reading it, though: because each release is a frozen, dated question set and the site re-runs old sets against new models, two “LiveBench” scores are only comparable when they share a release — a model’s number can rise or fall simply because the exam changed. Because each refresh changes the underlying questions, LiveBench functions less as a fixed yardstick than as a standing methodology for building one — a response to benchmark contamination rather than a single benchmark in the usual sense.