半价干翻Fable 5？Opus 5实测炸场，网友：差点从椅子上摔下来
Ranking
No observed public metrics; popularity remains neutral/archived.
Merged summary
TL;DR - Anthropic released Claude Opus 5, reporting major gains in coding, tool use, novel-problem solving, and computer operation at half the price of Fable 5. Early demonstrations also highlight stronger self-verification and enough model judgment for Claude Code to cut its system prompt by over 80% without benchmark regression.
- Opus 5 costs $5 per million input tokens and $25 per million output tokens, unchanged from Opus 4.8.
- It scored 43.3% on Frontier-Bench, 30.2% on ARC-AGI-3, and 64.7% on tool-enabled Humanity’s Last Exam.
- User tests showed strong generation of interactive games, simulations, CAD models, and single-file web experiences, often with accurate physics.
- The model increasingly builds its own validation pipelines, measuring outputs against source data rather than relying only on visual inspection.
Sources (1)
半价干翻Fable 5？Opus 5实测炸场，网友：差点从椅子上摔下来
TL;DR - Anthropic released Claude Opus 5, reporting major gains in coding, tool use, novel-problem solving, and computer operation at half the price of Fable 5. Early demonstrations also highlight stronger self-verification and enough model judgment for Claude Code to cut its system prompt by over 80% without benchmark regression.
- Opus 5 costs $5 per million input tokens and $25 per million output tokens, unchanged from Opus 4.8.
- It scored 43.3% on Frontier-Bench, 30.2% on ARC-AGI-3, and 64.7% on tool-enabled Humanity’s Last Exam.
- User tests showed strong generation of interactive games, simulations, CAD models, and single-file web experiences, often with accurate physics.
- The model increasingly builds its own validation pipelines, measuring outputs against source data rather than relying only on visual inspection.