Claude Fable 5 ranks 10th in document-writing benchmark, costs more too
A developer tested 30 AI models on 8 identical PowerPoint and Word document tasks, using the same minimal agent setup, environment, and document skills for every model. The finished files were ranked through blind human comparisons, gathering over 4,000 votes from 313 distinct voters. landed at #10, costing roughly $1.07 per document.
Every model ranked above it costs less: Qwen3.7 Plus is #3 at $0.07, MiniMax-M3 is #5 at $0.17, Grok 4.5 is #6 at $0.22, and is #7 at $0.56. Even Anthropic's own outranks Fable 5 while costing only $0.84. Because of its , Fable 5 completed only 7 of the 8 tasks, while Opus completed all 8.
Plotted on a quality-versus-cost chart, Fable 5 is dominated by nine other models that are both cheaper and preferred more often by voters. The tester notes the underlying benchmark, DocBench Arena, recently added more models, so comparisons involving Fable 5 are still based on limited data and rankings could shift significantly as more votes come in.
Key points
- 30 AI models tested on 8 identical PowerPoint/Word document tasks
- Results ranked via blind human comparison: 4,000+ votes from 313 voters
- ranked #10, costing about $1.07 per document — pricier than every model above it
- caused Fable 5 to complete only 7 of 8 tasks, versus 8/8 for
- The benchmark (DocBench Arena) has limited vote data so far, so rankings may shift