GDPval-AA normalized (GDPval-AA)
An Artificial Analysis normalized score for economically valuable tasks.
Top models on GDPval-AA — October 29, 2026
As of October 29, 2026, Claude Opus 5.5 leads the GDPval-AA leaderboard with 68.3% , followed by Muse Wonder Full (67.9%) and Kimi K3.1 Pro (67.3%).
Claude Opus 5.5
Anthropic
Muse Wonder Full
Meta
Kimi K3.1 Pro
Moonshot AI
143 modelsAgentic8% of Agentic reference weightCurrentUpdated October 29, 2026
Leaderboard (143 models)
| Rank | Model / configuration | Score | Parameters (B) | Open / closed |
|---|---|---|---|---|
| 1 | Claude Opus 5.5Anthropic | 68.3% | Not reported | Closed |
| 2 | Muse Wonder FullMeta | 67.9% | Not reported | Closed |
| 3 | Kimi K3.1 ProMoonshot AI | 67.3% | Not reported | Pending |
| 4 | Claude Sonnet 5.5Anthropic | 66.9% | Not reported | Closed |
| 5 | Muse WonderMeta | 63.5% | Not reported | Closed |
| 6 | Claude Fable 5.1Anthropic | 62.7% | Not reported | Closed |
| 7 | Claude Opus 5Anthropic | 61.1% | Not reported | Closed |
| 8 | Grok 4.7xAI | 60.7% | Not reported | Closed |
| 9 | MiMo-V2.6-ProXiaomi | 59.2% | Not reported | Open |
| 10 | Muse Spark 1.3Meta | 59.0% | Not reported | Closed |
| 11 | Qwen3.8 Max PreviewAlibaba | 58.3% | Not reported | Closed |
| 12 | GLM-5.3Z.AI | 57.5% | Not reported | Open |
| 13 | Qwen3.8-Flash-NextAlibaba | 56.4% | Not reported | Open |
| 14 | Gemini 4 ArgonGoogle | 56.2% | Not reported | Closed |
| 15 | Claude Haiku 5.5Anthropic | 56.0% | Not reported | Closed |
| 16 | Ling 3.1 FlashInclusionAI | 55.9% | Not reported | Open |
| 17 | Grok 4.6xAI | 55.9% | Not reported | Closed |
| 18 | Claude Fable 5Anthropic | 55.5% | Not reported | Closed |
| 19 | GPT-5.6 SolOpenAI | 55.4% | Not reported | Closed |
| 20 | MiMo-V2.6-FlashXiaomi | 55.4% | Not reported | Open |
| 21 | DeepSeek V4.1 Flash 1017DeepSeek | 55.0% | Not reported | Open |
| 22 | GPT-6.1 SolOpenAI | 54.8% | Not reported | Closed |
| 23 | DeepSeek V4 Pro 0813DeepSeek | 54.5% | Not reported | Open |
| 24 | Step 5 PreviewStepFun | 54.3% | Not reported | Open |
| 25 | GPT-6 AstraOpenAI | 53.7% | Not reported | Closed |
| 26 | Kimi K3Moonshot AI | 51.6% | Not reported | Open |
| 27 | GPT-6 SolOpenAI | 50.4% | Not reported | Closed |
| 28 | Muse Spark 1.2Meta | 48.6% | Not reported | Closed |
| 29 | Claude Sonnet 5Anthropic | 48.2% | Not reported | Closed |
| 30 | GPT-5.6 LunaOpenAI | 48.0% | Not reported | Closed |
| 31 | MiniMax M3.1 FlashMiniMax | 47.9% | Not reported | Closed |
| 32 | Claude Opus 4.8Anthropic | 47.7% | Not reported | Closed |
| 33 | GPT-5.6 TerraOpenAI | 47.6% | Not reported | Closed |
| 34 | Gemini 3.8 FlashGoogle | 46.8% | Not reported | Closed |
| 35 | DeepSeek V4 Flash 0731DeepSeek | 46.7% | Not reported | Open |
| 36 | GPT-6 LunaOpenAI | 46.6% | Not reported | Closed |
| 37 | Mistral Large 4Mistral | 46.1% | Not reported | Open |
| 38 | Qwen3.8-27BAlibaba | 45.9% | Not reported | Open |
| 39 | Muse Spark OpenMeta | 45.1% | Not reported | Open |
| 40 | Gemini 3.7 FlashGoogle | 44.6% | Not reported | Closed |
| 41 | Grok 4.5xAI | 44.2% | Not reported | Closed |
| 42 | GLM-5.2Z.AI | 43.6% | Not reported | Open |
| 43 | Claude Opus 4.7 (Adaptive)Anthropic | 42.6% | Not reported | Closed |
| 44 | GPT-5.5OpenAI | 42.5% | Not reported | Closed |
| 45 | Gemini 3.5 FlashGoogle | 42.2% | Not reported | Closed |
| 46 | DeepSeek V4 Pro (High)DeepSeek | 39.7% | Not reported | Open |
| 47 | Gemini 3.6 FlashGoogle | 39.2% | Not reported | Closed |
| 48 | GPT-5.4OpenAI | 37.3% | Not reported | Closed |
| 49 | MiniMax M3MiniMax | 37.1% | Not reported | Open |
| 50 | Hy3 PreviewTencent | 35.8% | Not reported | Open |
| 51 | Muse Spark 1.1Meta | 35.6% | Not reported | Closed |
| 52 | Apodex 1.1Apodex | 34.4% | Not reported | Closed |
| 53 | Apodex 1.1 MiniApodex | 34.4% | Not reported | Open |
| 54 | Quasar 438BMultiverse Computing | 34.1% | Not reported | Closed |
| 55 | Beam 1.0 501B-A23BReflection AI | 33.9% | Not reported | Open |
| 56 | Nemotron 3 UltraNVIDIA | 33.1% | Not reported | Open |
| 57 | Nemotron 3.5 Super VL 120B A12BNVIDIA | 33.1% | Not reported | Open |
| 58 | Nemotron 3.5 Super VL 120B A12B NVFP4NVIDIA | 33.1% | Not reported | Open |
| 59 | Ling 3.0 Flash VLInclusionAI | 33.0% | Not reported | Open |
| 60 | Qwen3.7 MaxAlibaba | 31.4% | Not reported | Closed |
| 61 | Inkling-SmallThinking Machines Lab | 31.2% | Not reported | Open |
| 62 | MiMo-V2.5-ProXiaomi | 31.0% | Not reported | Closed |
| 63 | GLM-5.1Z.AI | 30.8% | Not reported | Open |
| 64 | Solar Pro 4Upstage | 30.3% | Not reported | Closed |
| 65 | Grok 4.3xAI | 29.2% | Not reported | Closed |
| 66 | InklingThinking Machines Lab | 28.6% | Not reported | Open |
| 67 | Hy3Tencent | 28.2% | Not reported | Open |
| 68 | Kimi K2.6Moonshot AI | 26.9% | Not reported | Open |
| 69 | Kimi K2.7 CodeMoonshot AI | 26.9% | Not reported | Open |
| 70 | Step 3.7 FlashStepFun | 25.8% | Not reported | Open |
| 71 | GLM-4.7Z.AI | 25.7% | Not reported | Open |
| 72 | GPT-5.4 miniOpenAI | 25.6% | Not reported | Closed |
| 73 | MiniMax M2.7MiniMax | 25.5% | Not reported | Open |
| 74 | Muse SparkMeta | 24.9% | Not reported | Closed |
| 75 | Qwen3.6 PlusAlibaba | 24.5% | Not reported | Closed |
| 76 | Qwen3.6-27BAlibaba | 24.3% | Not reported | Open |
| 77 | Gemini 3.5 Flash-LiteGoogle | 24.2% | Not reported | Closed |
| 78 | Nemotron 3.5 Bolt 74B-A9BNVIDIA | 24.0% | Not reported | Open |
| 79 | A.X K2SK Telecom | 23.0% | Not reported | Open |
| 80 | Ling 3.0 FlashInclusionAI | 22.4% | Not reported | Open |
| 81 | GPT-5.4 nanoOpenAI | 22.4% | Not reported | Closed |
| 82 | GPT-5 (high)OpenAI | 20.9% | Not reported | Closed |
| 83 | Qwen3.6-35B-A3BAlibaba | 19.6% | Not reported | Open |
| 84 | Kimi K2.5 (Reasoning)Moonshot AI | 17.0% | Not reported | Closed |
| 85 | Kimi K2.5Moonshot AI | 17.0% | Not reported | Open |
| 86 | GPT-5.1OpenAI | 16.3% | Not reported | Closed |
| 87 | Qwen3.5-122B-A10BAlibaba | 15.7% | Not reported | Open |
| 88 | K-EXAONE 2.0LG AI Research | 15.4% | Not reported | Open |
| 89 | Gemini 3.1 ProGoogle | 14.6% | Not reported | Closed |
| 90 | Muse Glimmer 30BMeta | 14.2% | Not reported | Open |
| 91 | Qwen3.7 PlusAlibaba | 13.3% | Not reported | Closed |
| 92 | GPT-5 miniOpenAI | 13.3% | Not reported | Closed |
| 93 | Mistral Medium 3.5 128BMistral | 13.0% | Not reported | Open |
| 94 | MiniCPM5-2BOpenBMB | 10.4% | Not reported | Open |
| 95 | Nemotron 3.5 Lightning 30B A3B NVFP4NVIDIA | 6.9% | Not reported | Open |
| 96 | MiMo-V2-FlashXiaomi | 6.4% | Not reported | Open |
| 97 | Gemma 4 31BGoogle | 5.9% | Not reported | Open |
| 98 | GPT-OSS 120BOpenAI | 5.4% | Not reported | Open |
| 99 | Granite 4.2 30BIBM | 3.7% | Not reported | Open |
| 100 | Gemma 4 26B A4BGoogle | 3.2% | Not reported | Open |
| 101 | Ling 3.0 TinyInclusionAI | 3.0% | Not reported | Open |
| 102 | Command A+Cohere | 0.7% | Not reported | Open |
| 103 | Gemma 4 12BGoogle | 0.0% | Not reported | Open |
| 104 | Gemini 2.5 ProGoogle | 0.0% | Not reported | Closed |
| 105 | DeepSeek V3 0324DeepSeek | 0.0% | Not reported | Open |
| 106 | Nemotron 3 Nano Omni 30B A3BNVIDIA | 0.0% | Not reported | Open |
| 107 | GPT-4oOpenAI | 0.0% | Not reported | Closed |
| 108 | Gemma 4 E4BGoogle | 0.0% | Not reported | Open |
| 109 | Gemma 4 E2BGoogle | 0.0% | Not reported | Open |
| 110 | Llama 3.3 70BMeta | 0.0% | Not reported | Open |
| 111 | Celeris-1Celeris | 0.0% | Not reported | Closed |
| 112 | Llama 4 MaverickMeta | 0.0% | Not reported | Open |
| 113 | Mercury 2.5Inception | 0.0% | Not reported | Closed |
| 114 | Mistral Small 4Mistral | 0.0% | Not reported | Open |
| 115 | Solar Pro 3Upstage | 0.0% | Not reported | Closed |
| 116 | Trinity-Large-ThinkingArcee AI | 0.0% | Not reported | Open |
| 117 | Granite 4.2 8BIBM | 0.0% | Not reported | Open |
| 118 | Granite 4.2 3BIBM | 0.0% | Not reported | Open |
| 119 | K-ExaoneLG AI Research | 0.0% | Not reported | Closed |
| 120 | Mistral Large 3Mistral | 0.0% | Not reported | Closed |
| 121 | Nemotron 3 Nano 30BNVIDIA | 0.0% | Not reported | Open |
| 122 | GPT-OSS 20BOpenAI | 0.0% | Not reported | Open |
| 123 | Llama 4 ScoutMeta | 0.0% | Not reported | Open |
| 124 | GPT-4.1 miniOpenAI | 0.0% | Not reported | Closed |
| 125 | GPT-4o miniOpenAI | 0.0% | Not reported | Closed |
| 126 | Ling 2.6 FlashInclusionAI | 0.0% | Not reported | Open |
| 127 | LFM2.5-2.6BLiquidAI | 0.0% | Not reported | Open |
| 128 | GPT-4.1 nanoOpenAI | 0.0% | Not reported | Closed |
| 129 | Mistral Small 3.1Mistral | 0.0% | Not reported | Open |
| 130 | Claude 3.5 HaikuAnthropic | 0.0% | Not reported | Closed |
| 131 | DeepSeek V3DeepSeek | 0.0% | Not reported | Open |
| 132 | Gemma 3 27BGoogle | 0.0% | Not reported | Open |
| 133 | Trinity-Large-PreviewArcee AI | 0.0% | Not reported | Open |
| 134 | North Mini CodeCohere | 0.0% | Not reported | Open |
| 135 | Mistral Small 4 (Reasoning)Mistral | 0.0% | Not reported | Open |
| 136 | Ultravox v0.6 Llama 3.3 70BFixie AI | 0.0% | Not reported | Open |
| 137 | Mercury 2Inception | 0.0% | Not reported | Closed |
| 138 | Ministral 3 14B (Reasoning)Mistral | 0.0% | Not reported | Open |
| 139 | Ministral 3 14BMistral | 0.0% | Not reported | Open |
| 140 | Ministral 3 8B (Reasoning)Mistral | 0.0% | Not reported | Open |
| 141 | Ministral 3 8BMistral | 0.0% | Not reported | Open |
| 142 | Ministral 3 3B (Reasoning)Mistral | 0.0% | Not reported | Open |
| 143 | Ministral 3 3BMistral | 0.0% | Not reported | Open |
According to BenchLM.ai, Claude Opus 5.5 leads the GDPval-AA benchmark with a score of 68.3%, followed by Muse Wonder Full (67.9%) and Kimi K3.1 Pro (67.3%). The top models are clustered within 1.0 points, suggesting this benchmark is nearing saturation for frontier models.
143 models have been evaluated on GDPval-AA. The benchmark falls in the Agentic category. BenchAlign v5.8 gives GDPval-AA 8% of the Agentic reference weight, so it moves the Agentic leaderboard and the overall ranking. Reference weights are relative weights in the calibrated model, not fixed shares of a score.
About GDPval-AA
Year
2026
Tasks
Economically valuable tasks
Format
Normalized score
Difficulty
Professional agentic workflows
OpenRouter's Grok 4.3 benchmark card displays GDPval-AA as a normalized percentage. BenchLM stores it separately from the Elo-style GDPval-AA rows used in provider comparison tables.
Freshness and provenance
Version
GDPval-AA 2026
Refresh cadence
Quarterly
Staleness state
Current
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.
Questions
What does GDPval-AA measure?
An Artificial Analysis normalized score for economically valuable tasks.
Which model scores highest on GDPval-AA?
Claude Opus 5.5 by Anthropic currently leads with a score of 68.3% on GDPval-AA.
How many models are evaluated on GDPval-AA?
143 AI models have published results on GDPval-AA in the BenchLM catalog.
Know when it’s worth switching models
The model to choose, the cheaper alternative, and the release we would wait on.
Read a sample issueJoin 6,100+ subscribers.
One email each week. Unsubscribe anytime.