Benchmark profile
LiveBench
A frequently refreshed benchmark with objective scoring across reasoning, coding, agentic coding, mathematics, data analysis, language, and instruction following.
Data verifiedHow we show LiveBench
We mirror the official LiveBench 2026-06-25 release: 40 model variants across 23 objective tasks in 7 categories. The visible score is the mean of the seven category averages, matching the source leaderboard.
LiveBench refreshes its questions to limit contamination, but each row still carries a specific model version and reasoning-effort setting. We keep the mirror display only and preserve the category and cost fields in the snapshot instead of collapsing them into weighted model scores.
LiveBench overall score on LiveBench — June 25, 2026
BenchLM mirrors the published livebench overall score view for LiveBench. GPT-5.6 Sol (max effort) leads the public snapshot at 82.4% , followed by Claude Fable 5 (max effort) (80.8%) and GPT-5.5 (xhigh effort) (79.9%). BenchLM does not use these results to rank models overall.
GPT-5.6 Sol (max effort)
OpenAI
gpt-5.6-sol-max
Claude Fable 5 (max effort)
Anthropic
claude-fable-5-max-effort
GPT-5.5 (xhigh effort)
OpenAI
gpt-5.5-xhigh
LiveBench overall score table (40 models)
ScoreThe published LiveBench snapshot is tightly clustered at the top: GPT-5.6 Sol (max effort) sits at 82.4%, while the third row is only 2.5 points behind. The broader top-10 spread is 4.4 points, so many of the published scores sit in a relatively narrow band.
40 models have been evaluated on LiveBench. The benchmark falls in the External benchmark mirrors category. BenchLM tracks this category separately from its weighted global scoring system, so these results are best compared on the dedicated Korean benchmark views. LiveBench is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.
About LiveBench
Year
2024
Tasks
23 objective tasks across 7 categories
Format
Mean of category averages
Difficulty
Broad frontier-model evaluation
LiveBench rotates questions to limit contamination and scores answers against objective ground truth. We mirror the current release-specific table, including category scores and cost fields. The overall score is the mean of category averages. Model versions and reasoning-effort settings remain separate, and the mirror does not feed weighted rankings.
BenchLM freshness & provenance
Version
LiveBench 2024
Refresh cadence
Annual
Staleness state
Refreshing
Question availability
Public benchmark set
BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.
FAQ
What does LiveBench measure?
A frequently refreshed benchmark with objective scoring across reasoning, coding, agentic coding, mathematics, data analysis, language, and instruction following.
Which model leads the published LiveBench snapshot?
GPT-5.6 Sol (max effort) currently leads the published LiveBench snapshot with 82.4% livebench overall score. BenchLM shows this benchmark for display only and does not use it in overall rankings.
How many models are evaluated on LiveBench?
40 AI models are included in BenchLM's mirrored LiveBench snapshot, based on the public leaderboard captured on June 25, 2026.
The AI models change fast. We track them for you.
A weekly brief for engineers and researchers covering new models, ranking shifts, and pricing changes.
Free. No spam. Unsubscribe anytime.