Highest index score
64.6
Claude Fable 5.2 (Max, Default Fallback)AI model benchmarks
Compare independent benchmark scores and API prices to find the right models for your work.
Artificial Analysis via OpenRouter100 modelsSnapshot
Model comparison
Top 10 by intelligence indexLowest input price
$1.20
GPT-6.1 Terra (Max)Lowest output price
$6.00
GPT-6.1 Terra (Max)Summary cards and bars compare this top 10 only. Index scores are not percentages.
Cost efficiency
55 model configurationsIndex score vs. input price
Each point is a model configuration. Toward the upper left means a higher score for a lower input price.
- Anthropic
- OpenAI
- moonshotai
- Meta
- xAI
- tencent
- Qwen
- z-ai
- thinkingmachines
- xiaomi
- minimax
- DeepSeek
- stepfun
- upstage
- Mistral
- reflection
- nvidia
- cohere
- ibm-granite
- Input-price frontier
2 free models are outside this logarithmic scale. Switch to Linear to include them.
The dashed frontier connects models with a higher score than every cheaper model, including free models in that comparison. It considers input pricing only; output prices and your workload also affect total cost. Equal spacing on this logarithmic price scale represents equal price ratios.
45 models are omitted from the scatter because this index score or input price is not reported.
View chart data
| Model configuration | Intelligence index | Input price | Output price |
|---|---|---|---|
| Claude Fable 5.2 (Max, Default Fallback) | 64.6 | $16.00 | $80.00 |
| Claude Opus 5.5 (Max, Default Fallback) | 59.1 | $4.40 | $22.00 |
| GPT-6 Astra (Max) | 56.9 | $10.00 | $50.00 |
| GPT-6.1 Sol (Max) | 54.6 | $2.00 | $10.00 |
| Claude Sonnet 5.5 (Max, Default Fallback) | 54.5 | $2.20 | $11.00 |
| Qwen3.8 Max | 53.4 | Not reported | Not reported |
| GPT-6.1 Terra (Max) | 51.3 | $1.20 | $6.00 |
| Gemini 4 Argon (High) | 50.4 | $2.20 | $11.00 |
| Claude Fable 5.1 (Xhigh, Default Fallback) | 50 | $5.00 | $25.00 |
| Kimi K3.1 Pro | 46.5 | $2.00 | $10.00 |
| Claude Opus 5 (Max) | 44.2 | $5.50 | $27.50 |
| GPT-6 Sol (Max) | 44.2 | $2.00 | $10.00 |
| Muse Wonder (Max) | 43.6 | $4.00 | $20.00 |
| Muse Spark 1.3 (Max) | 36.6 | $1.25 | $4.25 |
| Claude Haiku 5.5 (Max) | 36.5 | $0.11 | $0.55 |
| Grok 4.7 (Xhigh) | 35.5 | $1.60 | $4.80 |
| Tencent Hy4 | 34.6 | $0.417 | $1.2505 |
| Qwen3.8 Max (0902) | 33.7 | $1.15 | $3.45 |
| GPT-5.6 Terra (Max) | 32.3 | $4.00 | $24.00 |
| GLM-5.3 (Max) | 31.7 | $0.573 | $1.169 |
| Grok 4.6 (Xhigh) | 31 | $1.60 | $4.80 |
| Gemini 3.8 Flash (High) | 30.8 | $0.825 | $4.125 |
| Inkling 2 Small | 30.6 | $0.00 | $0.00 |
| MiMo-V2.6-Pro | 29.9 | $4.00 | $8.00 |
| MiniMax-M3.1-Flash | 29.1 | $0.135 | $0.54 |
| DeepSeek V4.1 Flash 1017 (High) | 28.6 | $0.0825 | $0.2475 |
| Step 5 Preview | 28.5 | $1.00 | $2.70 |
| Solar Pro 4 | 28.2 | $0.15 | $0.60 |
| GPT-6 Luna (Max) | 27.9 | $0.10 | $0.50 |
| GLM-5.3-Flash | 27.8 | $0.15 | $0.50 |
| MiMo-V2.6-Flash | 26.7 | $0.14 | $0.28 |
| Mistral Large 4 (High) | 26.1 | $0.77 | $2.31 |
| Muse Spark Open (Xhigh) | 24.9 | $1.25 | $4.25 |
| GPT-5.6 Luna (Max) | 23.8 | $0.20 | $1.20 |
| DeepSeek V4 Pro 0813 (Max) | 23.1 | $1.98 | $5.94 |
| Qwen3.8 27B (Xhigh) | 20.1 | $0.16 | $2.00 |
| MiniMax-M3 | 18.9 | $0.224 | $0.88 |
| Beam 1.0 | 17.7 | $0.40 | $1.20 |
| Kimi K3 (Low) | 17.4 | $2.30 | $14.00 |
| Nemotron 3.5 Super VL | 13.6 | $0.15 | $0.50 |
| Inkling Small | 12.9 | $0.36 | $0.96 |
| DeepSeek V4 Pro 0813 (Non-reasoning) | 12.6 | $1.56 | $3.12 |
| Inkling (Xhigh) | 12.5 | $1.00 | $4.05 |
| Nemotron 3 Ultra 550B A55B (Reasoning) | 11.8 | $0.40 | $1.76 |
| Gemini 3.5 Flash-Lite | 11.7 | $0.54 | $4.50 |
| Claude 4.5 Haiku (Reasoning) | 7.4 | $1.10 | $5.50 |
| Muse Glimmer (High) | 6.3 | $0.24 | $0.96 |
| Command A+ | 6.2 | $0.30 | $1.50 |
| Gemini 3.1 Flash-Lite | 6.2 | $0.45 | $2.70 |
| Nemotron 3.5 Lightning | 5.5 | $0.05 | $0.20 |
| Granite 4.2 8B | 5.1 | $0.06 | $0.25 |
| Mistral Medium 3.5 | 4.6 | $0.75 | $3.75 |
| Gemma 4 31B (Reasoning) | 4.4 | $0.34 | $0.10 |
| Mistral Small 4 (Reasoning) | 4.3 | $0.165 | $0.66 |
| gpt-oss-120b (High) | 3.8 | $0.15 | $0.75 |
| North Mini Code | 3.4 | $0.00 | $0.00 |
Leaderboard
Explore benchmark scores and API prices. Shortlist up to three models to compare.
Top 10 of 100 models
Select 2–3 to compare
| Position# | Model | Intelligence index | API pricing / 1M tokens | Add to comparison |
|---|---|---|---|---|
| 1 | Claude Fable 5.2 (Max, Default Fallback) | Intelligence 64.6 | In $16.00 / Out $80.00 / 1M tokens | |
| 2 | Claude Opus 5.5 (Max, Default Fallback) | Intelligence 59.1 | In $4.40 / Out $22.00 / 1M tokens | |
| 3 | GPT-6 Astra (Max) | Intelligence 56.9 | In $10.00 / Out $50.00 / 1M tokens | |
| 4 | GPT-6.1 Sol (Max) | Intelligence 54.6 | In $2.00 / Out $10.00 / 1M tokens | |
| 5 | Claude Sonnet 5.5 (Max, Default Fallback) | Intelligence 54.5 | In $2.20 / Out $11.00 / 1M tokens | |
| 6 | Qwen3.8 Max | Intelligence 53.4 | In Not reported / Out Not reported / 1M tokens | |
| 7 | GPT-6.1 Terra (Max) | Intelligence 51.3 | In $1.20 / Out $6.00 / 1M tokens | |
| 8 | Gemini 4 Argon (High) | Intelligence 50.4 | In $2.20 / Out $11.00 / 1M tokens | |
| 9 | Claude Fable 5.1 (Xhigh, Default Fallback) | Intelligence 50.0 | In $5.00 / Out $25.00 / 1M tokens | |
| 10 | Kimi K3.1 Pro | Intelligence 46.5 | In $2.00 / Out $10.00 / 1M tokens |
Prices in USD. Missing values are “Not reported.” Configuration labels are preserved from the source.
Methodology & sources
Source: Artificial Analysis (artificialanalysis.ai) via OpenRouter (openrouter.ai/rankings).
Licensed under CC BY 4.0. View the license. Artificial Analysis source.
How to read these results
Each view uses one Artificial Analysis index. Intelligence, coding and agentic scores measure different evaluations. They are not percentages and are not combined with popularity or individual test accuracies. Configuration labels are preserved from the source.
Prices are USD per million input or output tokens. The cost-efficiency chart compares input prices only; output prices and your workload also affect total cost. Shortlist models to estimate token charges for the same workload. Actual costs depend on tokens consumed and provider rates.
These general benchmarks do not measure CRE underwriting accuracy. Test a shortlist on the same rent roll, T-12 and lease examples, checking extraction, arithmetic and cited evidence against known answers before using it for live work.
The snapshot timestamp identifies the source data snapshot, not when every model was evaluated.