reasoning benchmark
All tasks required; arithmetic mean: theory_of_mind, zebra_puzzle, spatial, logic_with_navigation
Updated Sep 25, 2026
Higher reasoning ranks better on this benchmark.
Rank | Model | reasoning | Percentile | Participants | Evidence | Evaluated |
|---|
| Rank01 | ModelOP | reasoning92.65 score | Percentile100.00% | Participants41 | EvidenceB | EvaluatedN/A |
| Rank02 | ModelOP | reasoning91.65 score | Percentile97.50% | Participants41 | EvidenceB | EvaluatedN/A |
| Rank03 | ModelMA | reasoning90.67 score | Percentile95.00% | Participants41 | EvidenceB | EvaluatedN/A |
| Rank04 | ModelOP | reasoning90.63 score | Percentile92.50% | Participants41 | EvidenceB | EvaluatedN/A |
| Rank05 | ModelXA | reasoning90.51 score | Percentile90.00% | Participants41 | EvidenceB | EvaluatedN/A |
| Rank06 | ModelME | reasoning90.00 score | Percentile87.50% | Participants41 | EvidenceB | EvaluatedN/A |
| Rank07 | ModelOP | reasoning89.65 score | Percentile85.00% | Participants41 | EvidenceB | EvaluatedN/A |
| Rank08 | ModelME | reasoning89.65 score | Percentile82.50% | Participants41 | EvidenceB | EvaluatedN/A |
| Rank09 | ModelGO | reasoning89.29 score | Percentile80.00% | Participants41 | EvidenceB | EvaluatedN/A |
| Rank10 | ModelAC | reasoning88.21 score | Percentile77.50% | Participants41 | EvidenceB | EvaluatedN/A |
| Rank11 | ModelOP | reasoning88.12 score | Percentile75.00% | Participants41 | EvidenceB | EvaluatedN/A |
| Rank12 | ModelGO | reasoning87.80 score | Percentile72.50% | Participants41 | EvidenceB | EvaluatedN/A |
| Rank13 | ModelME | reasoning87.73 score | Percentile70.00% | Participants41 | EvidenceB | EvaluatedN/A |
| Rank14 | ModelAC | reasoning87.38 score | Percentile67.50% | Participants41 | EvidenceB | EvaluatedN/A |
| Rank15 | ModelXA | reasoning87.17 score | Percentile65.00% | Participants41 | EvidenceB | EvaluatedN/A |
| Rank16 | ModelDE | reasoning86.63 score | Percentile62.50% | Participants41 | EvidenceB | EvaluatedN/A |
| Rank17 | ModelDE | reasoning85.84 score | Percentile60.00% | Participants41 | EvidenceB | EvaluatedN/A |
| Rank18 | ModelZA | reasoning85.80 score | Percentile57.50% | Participants41 | EvidenceB | EvaluatedN/A |
| Rank19 | ModelOP | reasoning85.64 score | Percentile55.00% | Participants41 | EvidenceB | EvaluatedN/A |
| Rank20 | ModelDE | reasoning85.40 score | Percentile52.50% | Participants41 | EvidenceB | EvaluatedN/A |
| Rank21 | ModelGO | reasoning85.15 score | Percentile50.00% | Participants41 | EvidenceB | EvaluatedN/A |
| Rank22 | ModelGO | reasoning84.00 score | Percentile47.50% | Participants41 | EvidenceB | EvaluatedN/A |
| Rank23 | ModelAC | reasoning83.34 score | Percentile45.00% | Participants41 | EvidenceB | EvaluatedN/A |
| Rank24 | ModelOP | reasoning83.21 score | Percentile42.50% | Participants41 | EvidenceB | EvaluatedN/A |
| Rank25 | ModelMA | reasoning82.81 score | Percentile40.00% | Participants41 | EvidenceB | EvaluatedN/A |
| Rank26 | ModelXA | reasoning82.65 score | Percentile37.50% | Participants41 | EvidenceB | EvaluatedN/A |
| Rank27 | ModelGO | reasoning82.00 score | Percentile35.00% | Participants41 | EvidenceB | EvaluatedN/A |
| Rank28 | ModelOP | reasoning81.10 score | Percentile32.50% | Participants41 | EvidenceB | EvaluatedN/A |
| Rank29 | ModelAC | reasoning80.03 score | Percentile30.00% | Participants41 | EvidenceB | EvaluatedN/A |
| Rank30 | ModelMA | reasoning79.38 score | Percentile27.50% | Participants41 | EvidenceB | EvaluatedN/A |
The leading models and scores on this benchmark.
A closer view of the leading scores on this benchmark.
The first five results on this benchmark, with official price and output speed added where the model identity can be matched.
What LiveBench reasoning (2026-06-25) measures and how its scores work.
All tasks required; arithmetic mean: theory_of_mind, zebra_puzzle, spatial, logic_with_navigation
Scores are shown in score. This benchmark is verified and has an evidence level of B.
LiveBench. Benchmark scores retain their original unit. Overall score eligibility is shown separately.
Common questions about LiveBench reasoning (2026-06-25).
GPT-6 Astra is currently ranked first with 92.65 score.
The current leaders are GPT-6 Astra (92.65 score), GPT-5.6 Sol (91.65 score), and Kimi K3 (90.67 score).
DeepSeek-V4-Flash-0731 has the lowest matched official input price at $0.15 input / $0.60 output per 1M tokens.
The fastest matched records are Gemini 3.6 Flash (188.24 tok/s via Google), Gemini 3.5 Flash-Lite (185.11 tok/s via Google), and Gemini 3.8 Flash (155.60 tok/s via Google).
No. This benchmark measures one defined capability or task. The overall LLMBoard score uses a separate aggregation across eligible benchmark evidence.
All tasks required; arithmetic mean: theory_of_mind, zebra_puzzle, spatial, logic_with_navigation
Yes. Higher values rank better for this benchmark.
41 model results are currently shown.
Yes. This benchmark can contribute to the current LLMBoard capability score.
Ranking basisThis livebench reasoning (2026-06-25) AI model leaderboard uses descending reasoning in the benchmark original unit. The leaderboard ranking keeps matched price and speed data separate from benchmark evidence.
Selection summary
GPT-6 Astra currently leads LiveBench reasoning (2026-06-25) with 92.65 score. It is the top model on this specific benchmark, while the best LLM for the broader task should also be checked against other benchmarks, price, and runtime.
Use this leaderboard with the supporting benchmark results and coverage details above. A leaderboard position summarizes the selected ranking signal; it does not replace workload-specific testing.