Artificial Analysis GDP.pdf (GDP.pdf)
We show this table for reference; we do not rank on it.
Artificial Analysis' document-output benchmark: professional tasks whose deliverable is a produced PDF, one of the ten components of its Intelligence Index v4.3.
Benchmark score on GDP.pdf — November 13, 2026
We compile the GDP.pdf rows from provider self-reports and secondary reports. Muse Wonder Full leads the table at 32.6%, followed by Kimi K3.1 Pro (31.4%) and GPT-6 Astra (31.0%). We do not use these results to rank models overall.
Muse Wonder Full
Meta
Carried from an upstream pageKimi K3.1 Pro
Moonshot AI
Carried from an upstream pageGPT-6 Astra
OpenAI
Carried from an upstream page22 modelsAgenticCurrentDisplay onlyUpdated November 13, 2026
Benchmark score table (22 models)
None of the 22 results on this page has a documented run by the benchmark or an independent evaluator: 1 reported by the lab that built the model and 21 carried from an upstream page, where who ran them is not recorded.
| Rank | Model / configuration | Score | Who ran it | Parameters (B) | Open / closed |
|---|---|---|---|---|---|
| 1 | Muse Wonder FullMetaCarried from an upstream page | 32.6% | Carried from an upstream page | Not reported | Closed |
| 2 | Kimi K3.1 ProMoonshot AICarried from an upstream page | 31.4% | Carried from an upstream page | Not reported | Open |
| 3 | GPT-6 AstraOpenAICarried from an upstream page | 31.0% | Carried from an upstream page | Not reported | Closed |
| 4 | GPT-6.1 SolOpenAICarried from an upstream page | 31.0% | Carried from an upstream page | Not reported | Closed |
| 5 | Muse WonderMetaCarried from an upstream page | 30.6% | Carried from an upstream page | Not reported | Closed |
| 6 | GPT-6.1 TerraOpenAICarried from an upstream page | 28.6% | Carried from an upstream page | Not reported | Closed |
| 7 | Gemini 4 CarbonGoogleCarried from an upstream page | 28.4% | Carried from an upstream page | Not reported | Closed |
| 8 | Gemini 4 HydrogenGoogleCarried from an upstream page | 27.6% | Carried from an upstream page | Not reported | Closed |
| 9 | Claude Fable 5.1AnthropicCarried from an upstream page | 26.2% | Carried from an upstream page | Not reported | Closed |
| 10 | Claude Opus 5.5AnthropicCarried from an upstream page | 26.2% | Carried from an upstream page | Not reported | Closed |
| 11 | Claude Sonnet 5.5AnthropicCarried from an upstream page | 25.8% | Carried from an upstream page | Not reported | Closed |
| 12 | Inkling 2 SmallThinking Machines LabCarried from an upstream page | 25.2% | Carried from an upstream page | Not reported | Closed |
| 13 | Claude Fable 5.2AnthropicCarried from an upstream page | 24.2% | Carried from an upstream page | Not reported | Closed |
| 14 | GPT-6 LunaOpenAICarried from an upstream page | 22.8% | Carried from an upstream page | Not reported | Closed |
| 15 | Muse EmberMetaReported by Meta | 22.8% | Reported by Meta | Not reported | Closed |
| 16 | Claude Haiku 5.5AnthropicCarried from an upstream page | 20.8% | Carried from an upstream page | Not reported | Closed |
| 17 | Grok 4.7xAICarried from an upstream page | 20.0% | Carried from an upstream page | Not reported | Closed |
| 18 | MiMo-V2.6-ProXiaomiCarried from an upstream page | 19.2% | Carried from an upstream page | Not reported | Open |
| 19 | Mistral Large 4MistralCarried from an upstream page | 17.6% | Carried from an upstream page | Not reported | Open |
| 20 | Qwen3.8-27BAlibabaCarried from an upstream page | 16.4% | Carried from an upstream page | Not reported | Open |
| 21 | Beam 1.0 501B-A23BReflection AICarried from an upstream page | 11.6% | Carried from an upstream page | Not reported | Open |
| 22 | GLM-5.3Z.AICarried from an upstream page | 11.2% | Carried from an upstream page | Not reported | Open |
- Lab-reported
- Benchmark-run
- Independent
Among the reported GDP.pdf rows, Muse Wonder Full is first at 32.6%. The third row is 1.6 points behind. The broader top-10 range is 6.4 points, so many of the published results sit in a relatively narrow band.
22 models have been evaluated on GDP.pdf. The benchmark falls in the Agentic category. GDP.pdf is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.
About GDP.pdf
Year
2026
Tasks
Professional document-production tasks
Format
Task success rate
Difficulty
Professional knowledge work
Independently run by Artificial Analysis under its published harness. Display-only; not yet admitted as independent-run evidence.
Freshness and provenance
Version
GDP.pdf 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 GDP.pdf measure?
Artificial Analysis' document-output benchmark: professional tasks whose deliverable is a produced PDF, one of the ten components of its Intelligence Index v4.3.
Which model scores highest on GDP.pdf?
Muse Wonder Full by Meta currently leads with a score of 32.6% on GDP.pdf.
How many models are evaluated on GDP.pdf?
22 AI models have published results on GDP.pdf in the BenchLM catalog.
Compare top models on GDP.pdf
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 7,000+ subscribers.
One email each week. Unsubscribe anytime.