GDPval-AA
We show this table for reference; we do not rank on it.
An agentic real-world work-task evaluation reported as an Elo score in DeepSeek-V4 thinking-mode evaluations.
Benchmark score on GDPval-AA — October 15, 2026
We compile the GDPval-AA rows from provider self-reports and secondary reports. Claude Opus 5 leads the table at 1862, followed by Muse Wonder Full (1857) and Claude Opus 5.5 (1846). We do not use these results to rank models overall.
Claude Opus 5
Anthropic
Muse Wonder Full
Meta
Claude Opus 5.5
Anthropic
124 modelsAgenticCurrentDisplay onlyUpdated October 15, 2026
Benchmark score table (124 models)
| Rank | Model / configuration | Score | Parameters (B) | Open / closed |
|---|---|---|---|---|
| 1 | Claude Opus 5Anthropic | 1862 | Not reported | Closed |
| 2 | Muse Wonder FullMeta | 1857 | Not reported | Closed |
| 3 | Claude Opus 5.5Anthropic | 1846 | Not reported | Closed |
| 4 | Claude Sonnet 5.5Anthropic | 1844 | Not reported | Closed |
| 5 | GLM-5.3-FlashZ.AI | 1773 | Not reported | Open |
| 6 | Muse WonderMeta | 1770 | Not reported | Closed |
| 7 | GLM-5.3Z.AI | 1769 | Not reported | Open |
| 8 | Claude Fable 5.1Anthropic | 1754 | Not reported | Closed |
| 9 | Muse Spark 1.3Meta | 1754 | Not reported | Closed |
| 10 | Claude Fable 5Anthropic | 1747 | Not reported | Closed |
| 11 | GPT-5.6 SolOpenAI | 1735 | Not reported | Closed |
| 12 | Grok 4.7xAI | 1711 | Not reported | Closed |
| 13 | Hy4 previewTencent | 1678 | Not reported | Open |
| 14 | Ling 3.1 FlashInclusionAI | 1673 | Not reported | Open |
| 15 | MiMo-V2.6-ProXiaomi | 1673 | Not reported | Open |
| 16 | Qwen3.8-Flash-NextAlibaba | 1648 | Not reported | Open |
| 17 | Grok 4.6xAI | 1643 | Not reported | Closed |
| 18 | Muse Spark 1.2Meta | 1631 | Not reported | Closed |
| 19 | Qwen3.8 Max PreviewAlibaba | 1630 | Not reported | Closed |
| 20 | Gemini 4 ArgonGoogle | 1624 | Not reported | Closed |
| 21 | Claude Haiku 5.5Anthropic | 1620 | Not reported | Closed |
| 22 | Claude Sonnet 5Anthropic | 1603 | Not reported | Closed |
| 23 | DeepSeek V4.1 FlashDeepSeek | 1600 | Not reported | Open |
| 24 | GPT-6.1 SolOpenAI | 1594 | Not reported | Closed |
| 25 | Claude Opus 4.8Anthropic | 1593 | Not reported | Closed |
| 26 | GPT-5.6 TerraOpenAI | 1583 | Not reported | Closed |
| 27 | Atria Dawn PreviewShanghai Artificial Intelligence Laboratory | 1583 | Not reported | Open |
| 28 | GPT-5.6 LunaOpenAI | 1582 | Not reported | Closed |
| 29 | GPT-6 AstraOpenAI | 1573 | Not reported | Closed |
| 30 | Step 5 PreviewStepFun | 1566 | Not reported | Open |
| 31 | Gemini 3.8 FlashGoogle | 1545 | Not reported | Closed |
| 32 | Kimi K3Moonshot AI | 1533 | Not reported | Pending |
| 33 | Gemini 3.7 FlashGoogle | 1525 | Not reported | Closed |
| 34 | GPT-6 LunaOpenAI | 1432 | Not reported | Closed |
| 35 | Grok 4.5xAI | 1430 | Not reported | Closed |
| 36 | Gemini 3.6 FlashGoogle | 1423 | Not reported | Closed |
| 37 | Mistral Large 4Mistral | 1422 | Not reported | Pending |
| 38 | Qwen3.8-27BAlibaba | 1420 | Not reported | Open |
| 39 | GLM-5.2Z.AI | 1418 | Not reported | Open |
| 40 | Muse Spark OpenMeta | 1401 | Not reported | Open |
| 41 | GPT-5.5OpenAI | 1396 | Not reported | Closed |
| 42 | Claude Opus 4.7 (Adaptive)Anthropic | 1396 | Not reported | Closed |
| 43 | Muse Spark 1.1Meta | 1375 | Not reported | Closed |
| 44 | Gemini 3.5 FlashGoogle | 1345 | Not reported | Closed |
| 45 | GPT-5.4OpenAI | 1307 | Not reported | Closed |
| 46 | DeepSeek V4 Pro 0813DeepSeek | 1306 | Not reported | Open |
| 47 | DeepSeek V4 Pro (High)DeepSeek | 1294 | Not reported | Open |
| 48 | Apodex 1.1Apodex | 1273 | Not reported | Closed |
| 49 | Apodex 1.1 MiniApodex | 1273 | Not reported | Open |
| 50 | MiMo-V2.5-ProXiaomi | 1265 | Not reported | Closed |
| 51 | MiniMax M3MiniMax | 1242 | Not reported | Open |
| 52 | Quasar 438BMultiverse Computing | 1239 | Not reported | Closed |
| 53 | Nemotron 3.5 Super VL 120B A12BNVIDIA | 1211 | Not reported | Open |
| 54 | Nemotron 3.5 Super VL 120B A12B NVFP4NVIDIA | 1206 | Not reported | Open |
| 55 | Inkling-SmallThinking Machines Lab | 1191 | Not reported | Open |
| 56 | Qwen3.7 MaxAlibaba | 1190 | Not reported | Closed |
| 57 | DeepSeek V4 Flash 0731DeepSeek | 1189 | Not reported | Open |
| 58 | GLM-5.1Z.AI | 1181 | Not reported | Open |
| 59 | Gemini 3.5 Flash-LiteGoogle | 1139 | Not reported | Closed |
| 60 | Hy3Tencent | 1136 | Not reported | Open |
| 61 | Hy3 PreviewTencent | 1136 | Not reported | Open |
| 62 | Kimi K2.6Moonshot AI | 1115 | Not reported | Open |
| 63 | Kimi K2.7 CodeMoonshot AI | 1114 | Not reported | Open |
| 64 | Ling 3.0 FlashInclusionAI | 1107 | Not reported | Open |
| 65 | GLM-4.7Z.AI | 1096 | Not reported | Open |
| 66 | GPT-5.4 miniOpenAI | 1095 | Not reported | Closed |
| 67 | Nemotron 3 UltraNVIDIA | 1091 | Not reported | Open |
| 68 | MiniMax M2.7MiniMax | 1087 | Not reported | Open |
| 69 | InklingThinking Machines Lab | 1076 | Not reported | Open |
| 70 | Muse SparkMeta | 1076 | Not reported | Closed |
| 71 | Qwen3.6-27BAlibaba | 1069 | Not reported | Open |
| 72 | Qwen3.6 PlusAlibaba | 1066 | Not reported | Closed |
| 73 | GPT-5.4 nanoOpenAI | 1035 | Not reported | Closed |
| 74 | Grok 4.3xAI | 1018 | Not reported | Closed |
| 75 | GPT-5 (high)OpenAI | 1015 | Not reported | Closed |
| 76 | Qwen3.6-35B-A3BAlibaba | 992 | Not reported | Open |
| 77 | Step 3.7 FlashStepFun | 954 | Not reported | Open |
| 78 | Kimi K2.5 (Reasoning)Moonshot AI | 936 | Not reported | Closed |
| 79 | Kimi K2.5Moonshot AI | 936 | Not reported | Open |
| 80 | GPT-5.1OpenAI | 930 | Not reported | Closed |
| 81 | Qwen3.5-122B-A10BAlibaba | 925 | Not reported | Open |
| 82 | Gemini 3.1 ProGoogle | 904 | Not reported | Closed |
| 83 | Qwen3.7 PlusAlibaba | 886 | Not reported | Closed |
| 84 | GPT-5 miniOpenAI | 880 | Not reported | Closed |
| 85 | Mistral Medium 3.5 128BMistral | 875 | Not reported | Open |
| 86 | Nemotron 3.5 Lightning 30B A3B NVFP4NVIDIA | 865 | Not reported | Open |
| 87 | MiMo-V2-FlashXiaomi | 788 | Not reported | Open |
| 88 | Muse Glimmer 30BMeta | 787 | Not reported | Open |
| 89 | Gemma 4 31BGoogle | 755 | Not reported | Open |
| 90 | GPT-OSS 120BOpenAI | 745 | Not reported | Open |
| 91 | Gemma 4 26B A4BGoogle | 713 | Not reported | Open |
| 92 | Command A+Cohere | 658 | Not reported | Open |
| 93 | Granite 4.2 8BIBM | 646 | Not reported | Open |
| 94 | Mercury 2Inception | 646 | Not reported | Closed |
| 95 | Gemini 2.5 ProGoogle | 616 | Not reported | Closed |
| 96 | Gemma 4 12BGoogle | 591 | Not reported | Open |
| 97 | Mistral Large 3Mistral | 588 | Not reported | Closed |
| 98 | Ling 2.6 FlashInclusionAI | 550 | Not reported | Open |
| 99 | K-ExaoneLG AI Research | 540 | Not reported | Closed |
| 100 | Mistral Small 4Mistral | 537 | Not reported | Open |
| 101 | Mistral Small 4 (Reasoning)Mistral | 537 | Not reported | Open |
| 102 | GPT-OSS 20BOpenAI | 514 | Not reported | Open |
| 103 | Trinity-Large-ThinkingArcee AI | 512 | Not reported | Open |
| 104 | Trinity-Large-PreviewArcee AI | 512 | Not reported | Open |
| 105 | Celeris-1Celeris | 483 | Not reported | Closed |
| 106 | GPT-4.1 miniOpenAI | 453 | Not reported | Closed |
| 107 | Nemotron 3 Nano 30BNVIDIA | 438 | Not reported | Open |
| 108 | Ministral 3 14B (Reasoning)Mistral | 432 | Not reported | Open |
| 109 | Ministral 3 14BMistral | 432 | Not reported | Open |
| 110 | Nemotron 3 Nano Omni 30B A3BNVIDIA | 416 | Not reported | Open |
| 111 | Ministral 3 8B (Reasoning)Mistral | 405 | Not reported | Open |
| 112 | Ministral 3 8BMistral | 405 | Not reported | Open |
| 113 | Ministral 3 3B (Reasoning)Mistral | 234 | Not reported | Open |
| 114 | Ministral 3 3BMistral | 234 | Not reported | Open |
| 115 | LFM2.5-2.6BLiquidAI | 202 | Not reported | Open |
| 116 | GPT-4o miniOpenAI | 188 | Not reported | Closed |
| 117 | DeepSeek V3DeepSeek | 183 | Not reported | Open |
| 118 | Gemma 4 E4BGoogle | 177 | Not reported | Open |
| 119 | Llama 4 ScoutMeta | 60 | Not reported | Open |
| 120 | Ultravox v0.6 Llama 3.3 70BFixie AI | 50 | Not reported | Open |
| 121 | Gemma 4 E2BGoogle | 36 | Not reported | Open |
| 122 | GPT-4.1 nanoOpenAI | 13 | Not reported | Closed |
| 123 | Llama 4 MaverickMeta | -45 | Not reported | Open |
| 124 | Gemma 3 27BGoogle | -171 | Not reported | Open |
Among the reported GDPval-AA rows, Claude Opus 5 is first at 1862. The third row is 16 score units behind. The broader top-10 range is 115 score units, so the table still separates the published systems.
124 models have been evaluated on GDPval-AA. The benchmark falls in the Agentic category. GDPval-AA is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.
About GDPval-AA
Year
2026
Tasks
Agentic real-world work tasks
Format
Elo
Difficulty
Professional agentic workflows
BenchLM stores GDPval-AA as a display-only provider-table row for DeepSeek-V4 because the source reports an Elo score rather than a 0-100 percentage.
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 agentic real-world work-task evaluation reported as an Elo score in DeepSeek-V4 thinking-mode evaluations.
Which model scores highest on GDPval-AA?
Claude Opus 5 by Anthropic currently leads with a score of 1862 on GDPval-AA.
How many models are evaluated on GDPval-AA?
124 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 5,500+ readers.
One email each week. Unsubscribe anytime.