All Benchmarks/Other

LiveBench 2.0 Design

Higher is better
What does it measure?
Tests whether a model can produce or rework a finished Word document, PDF or slide deck to a brief.
Who evaluates it, and how?
Tasks include restyling documents, PDFs such as annual reports and posters, and building or rebranding slide decks (25 tasks in the October 25, 2026 release). The produced file is rendered and scored on visual and structural checks against the brief.
Reading the score
Scores are shown as published by the source. Higher is better.
Other versions:LiveBench 2.0LiveBench 2.0 BrowserLiveBench 2.0 CodingLiveBench 2.0 Data AnalysisLiveBench 2.0 Long AgenticLiveBench 2.0 ReasoningLiveBench 2.0 Research
Top score85.1 / 100Claude Fable 5.2
Models tested22
Last updated2026.11.06
Model release date (newest on the right)
Best score so farHigher on the chart is better
  1. 1
    Anthropic
    Claude Fable 5.2
    2026.11.09 · xhigh
    85.1 / 100
  2. 2
    Anthropic
    Claude Opus 5.5
    2026.09.22 · max
    83.3 / 100
  3. 3
    Anthropic
    Claude Fable 5.1
    2026.09.01 · max
    81.6 / 100
  4. 4
    Anthropic
    Claude Sonnet 5.5
    2026.09.28 · max
    81.3 / 100
  5. 5
    OpenAI
    GPT-6 Astra
    2026.09.04 · max
    80.0 / 100
  6. 6
    DeepSeek
    DeepSeek V4.1 Flash 1017
    2026.10.17 · max
    79.8 / 100
  7. 7
    OpenAI
    GPT-6.1 Terra
    2026.10.26 · max
    79.7 / 100
  8. 8
    OpenAI
    GPT-6.1 Sol
    2026.09.29 · max
    79.6 / 100
  9. 9
    Anthropic
    Claude Haiku 5.5
    2026.10.07 · max
    79.3 / 100
  10. 10
    SpaceXSI
    Grok 4.6
    2026.08.12 · xhigh
    76.3 / 100

The "harness" is the agent program the AI used to carry out the task. The same model can score very differently depending on its harness and reasoning effort.

Where the scores come from

  • Run by the test’s team: scores the team that operates the test ran and published itself.