OPEN BENCHMARKS
AI Model Benchmarks
Compare language, image, video and speech models with traceable open benchmark data.
Compare models
0/3Choose 2–3 models from this leaderboard. Only measurements with the same source, version and track are compared.
| Rank | Model | Score | Creator | Source |
|---|---|---|---|---|
| 1 | claude-fable-5-1-max-effort | 83.783 score | Anthropic | LiveBench |
| 2 | claude-fable-5-max-effort | 83.383 score | Anthropic | LiveBench |
| 3 | claude-opus-5-5-max-effort | 83.185 score | Anthropic | LiveBench |
| 4 | gpt-6-astra-max | 83.046 score | OpenAI | LiveBench |
| 5 | gpt-6.1-sol-max | 82.487 score | OpenAI | LiveBench |
| 6 | muse-spark-1.3-xhigh | 82.431 score | Unknown | LiveBench |
| 7 | claude-opus-5-5-xhigh-effort | 82.109 score | Anthropic | LiveBench |
| 8 | gpt-6.1-sol-xhigh | 81.827 score | OpenAI | LiveBench |
| 9 | gpt-5.6-sol-max | 81.647 score | OpenAI | LiveBench |
| 10 | deepseek-v4.1-flash-max | 81.447 score | DeepSeek | LiveBench |
| 11 | gpt-5.5-xhigh | 80.785 score | OpenAI | LiveBench |
| 12 | claude-opus-5-max-effort | 80.48 score | Anthropic | LiveBench |
| 13 | gemini-3.7-flash-high | 79.898 score | LiveBench | |
| 14 | gpt-6-sol-max | 79.825 score | OpenAI | LiveBench |
| 15 | smaug-agentic | 79.669 score | Unknown | LiveBench |
| 16 | qwen3.8-max | 79.496 score | Alibaba | LiveBench |
| 17 | kimi-k3 | 79.482 score | Moonshot AI | LiveBench |
| 18 | grok-4.6 | 79.002 score | xAI | LiveBench |
| 19 | muse-spark-1.2-xhigh | 78.915 score | Unknown | LiveBench |
| 20 | gpt-5.4-xhigh | 78.798 score | OpenAI | LiveBench |
| 21 | gpt-5.6-terra-max | 78.634 score | OpenAI | LiveBench |
| 22 | grok-4.7-xhigh | 78.338 score | xAI | LiveBench |
| 23 | claude-sonnet-5-5-xhigh-effort | 78.174 score | Anthropic | LiveBench |
| 24 | deepseek-v4-pro-0813 | 78.158 score | DeepSeek | LiveBench |
| 25 | smaug-flash | 78.026 score | Unknown | LiveBench |
| 26 | gemini-3.1-pro-preview-high | 77.986 score | LiveBench | |
| 27 | deepseek-v4-flash-vision-exp | 77.734 score | DeepSeek | LiveBench |
| 28 | smaug-mini | 77.678 score | Unknown | LiveBench |
| 29 | gemini-3.8-flash-high | 77.486 score | LiveBench | |
| 30 | qwen3.8-flash-next | 77.297 score | Alibaba | LiveBench |
| 31 | claude-opus-4-8-max-effort | 77.149 score | Anthropic | LiveBench |
| 32 | grok-4.5 | 77.051 score | xAI | LiveBench |
| 33 | claude-opus-4-7-xhigh-effort | 77.035 score | Anthropic | LiveBench |
| 34 | glm-5.3 | 76.65 score | Unknown | LiveBench |
| 35 | claude-sonnet-5-xhigh-effort | 76.593 score | Anthropic | LiveBench |
| 36 | muse-spark-1.1-xhigh | 76.028 score | Unknown | LiveBench |
| 37 | qwen3.8-27b | 75.82 score | Alibaba | LiveBench |
| 38 | claude-sonnet-5-5-max-effort | 75.752 score | Anthropic | LiveBench |
| 39 | gemini-3.5-flash-high | 75.437 score | LiveBench | |
| 40 | gpt-5.2-2025-12-11-high | 75.192 score | OpenAI | LiveBench |
| 41 | claude-opus-4-6-thinking-auto-high-effort | 75.132 score | Anthropic | LiveBench |
| 42 | deepseek-v4-flash-0731 | 74.85 score | DeepSeek | LiveBench |
| 43 | gemini-3.6-flash-high | 74.539 score | LiveBench | |
| 44 | qwen3.7-max | 74.099 score | Alibaba | LiveBench |
| 45 | gpt-5.2-codex | 74.034 score | OpenAI | LiveBench |
| 46 | gpt-5.6-luna-max | 73.687 score | OpenAI | LiveBench |
| 47 | claude-sonnet-4-6-thinking-auto-medium-effort | 73.413 score | Anthropic | LiveBench |
| 48 | glm-5.2 | 73.363 score | Unknown | LiveBench |
| 49 | claude-opus-4-5-20251101-thinking-64k-high-effort | 72.939 score | Anthropic | LiveBench |
| 50 | inkling-xhigh | 72.876 score | Unknown | LiveBench |
| 51 | deepseek-v4-pro | 72.554 score | DeepSeek | LiveBench |
| 52 | gpt-6-luna-max | 72.193 score | OpenAI | LiveBench |
| 53 | glm-5.3-flash | 71.138 score | Unknown | LiveBench |
| 54 | kimi-k2.6-thinking | 70.902 score | Moonshot AI | LiveBench |
| 55 | gpt-5.4-nano-xhigh | 70.85 score | OpenAI | LiveBench |
| 56 | ox-alpha-max | 69.242 score | Unknown | LiveBench |
| 57 | qwen3.6-plus | 68.989 score | Alibaba | LiveBench |
| 58 | kimi-k2.7-code | 68.756 score | Moonshot AI | LiveBench |
| 59 | nemotron-3-ultra-550b-a55b | 68.723 score | Unknown | LiveBench |
| 60 | grok-build-0.1 | 68.61 score | xAI | LiveBench |
| 61 | minimax-m3 | 67.527 score | Unknown | LiveBench |
| 62 | gpt-5.4-mini-xhigh | 66.601 score | OpenAI | LiveBench |
| 63 | deepseek-v4-flash | 66.054 score | DeepSeek | LiveBench |
| 64 | qwen3.6-27b | 64.181 score | Alibaba | LiveBench |
| 65 | gemini-3.5-flash-lite-high | 63.816 score | LiveBench | |
| 66 | grok-4.3 | 63.268 score | xAI | LiveBench |
No matching measured models.
EXTERNAL REFERENCE
More from Artificial Analysis
Explore these dimensions beyond benchmark scores. Follow the official pages for current data.
- Intelligence IndexCombines independent evaluations; higher index scores mean stronger overall results.
- Coding agentsCompare end-to-end software task results, cost and execution time.
- Capability indexesExplore specialist indexes for finance, law, healthcare and engineering.
- Speed and latencyDistinguish output tokens/s, first-token delay and full response time.
- Pricing and task costDistinguish per-million-token prices from the cost of an evaluation task.
- Context and token useCheck both context-window size and output tokens used per task.
- OpennessCompare open weights, usage restrictions and model transparency.
- Multilingual abilityCompare languages with a separate index; it is not part of the mainly English, text-based Intelligence Index.
- Long-document reasoningUse AA-LCR to assess reasoning across long documents; a large context window alone does not prove accuracy.
- API providersCompare price, output speed and first-response latency across endpoints serving the same model.
- Knowledge-work benchmarkExplore AA-Briefcase task results, analytical quality and presentation quality for deliverables such as spreadsheets, slides and memos.
- Release comparisonCompare model releases and their reasoning variants side by side on capability, cost, speed and context.
- Model recommenderGet official model suggestions weighted by your priorities for intelligence, speed and cost.
Focused evaluations
Explore all evaluations ↗- GDPval-AAAssess tool use and file deliverables in real professional tasks, with blind comparisons of outcomes.
- AutomationBench-AAMeasure goal completion in simulated SaaS workflows while accounting for guardrail violations.
- AA-OmniscienceSeparate correct factual answers, hallucinations and appropriate abstentions.
- GDP.pdfAssess reading and reasoning over long professional PDFs against specific task requirements.
- IFBenchTest adherence to novel, verifiable output constraints.
Sources and methodology
LiveBenchData is older
- Version
- 2026-06-25
- License
- Apache-2.0
- Updated
- 2026-06-25T00:00:00.000Z