Claude Opus 5.5 Matches Fable 5.1 Performance at 40% Lower Cost

Claude Opus 5.5, launched on September 22, 2026, offers a 40% reduction in operating costs compared to its predecessor, Opus 5. The new model charges $4 per million tokens for input and $20 for output, down from $5 and $25 respectively for Opus 5. Performance benchmarks show Opus 5.5 achieving 66.4% on Terminal-Bench 4.0, surpassing Fable 5.1's 55.8%. Additionally, it scores 67.7% on Humanity's Last Exam, compared to Fable's 65.6%. On AutomationBench, Opus 5.5 reaches 40.0%, while Fable 5.1 only achieves 31.4%. Opus 5.5 also outperforms GPT-5.6 Sol on CursorBench with a score of 57.8% versus 46.6%.

The model's Max Effort Score on the Artificial Analysis Index is 58, improving from Fable 5.1's previous score of 53. Opus 5.5 uses 119,000 output tokens per task, more than Fable 5.1's 78,000 and Opus 5's 73,000, but significantly less than GPT-6 Astra's 27,000. Anthropic claims that Opus 5.5 communicates more naturally than earlier models, addressing previous concerns about