Anthropic Launches Claude Opus 5: State-of-the-Art Performance at Half the Cost of Fable 5
Key Takeaways
- ▸Claude Opus 5 is now the default model on Claude Max, achieving state-of-the-art performance on coding and knowledge work benchmarks while costing the same as Opus 4.8
- ▸The model performs within 0.5% of Fable 5's peak performance while costing approximately half as much, positioning it as a compelling value option
- ▸Opus 5 demonstrates 3x performance improvement on novel problem-solving (ARC-AGI) and 1.5x pass rates on business automation tasks compared to competing models
Summary
Anthropic has announced Claude Opus 5, a new large language model that delivers near-frontier performance at significantly lower cost than its flagship Fable 5 model. Available today as the new default model on Claude Max and the strongest model on Claude Pro, Opus 5 represents a major upgrade over its predecessor Opus 4.8 while maintaining the same cost.
The model demonstrates exceptional performance across multiple evaluation benchmarks. On coding tasks (Frontier-Bench v0.1, CursorBench 3.2), Opus 5 achieves state-of-the-art results, more than doubling Opus 4.8's performance at lower cost per task. On knowledge work and reasoning tasks like ARC-AGI and Zapier AutomationBench, it significantly outperforms competing models—achieving 3x the score on ARC-AGI and 1.5x the pass rate on Zapier. On computer use benchmarks (OSWorld 2.0), it surpasses Fable 5's results at roughly one-third of the cost.
Beyond benchmarks, Opus 5 demonstrates enhanced reasoning and agency in real-world scenarios. The model successfully wrote a computer vision pipeline to interpret machine part drawings when direct image access was blocked, debugged subtle edge cases in open-source code that competitors missed, and enabled a trading firm engineer to build a complex market data feed in a single session—a task previous models could not complete even with extensive planning. The model excels particularly in scientific research, showing significant improvements in organic chemistry and protein-related tasks.
- Real-world testing shows enhanced reasoning and iterative capability, including solving tasks like custom computer vision pipelines and finding deep bugs that other models missed
Editorial Opinion
Claude Opus 5 signals a maturation in Anthropic's product strategy—closing the capability-to-cost gap that many developers have been waiting for. By delivering frontier-adjacent performance at mid-tier pricing, Anthropic is effectively raising the baseline quality for production AI work. The demonstrated agency and reasoning depth, particularly in novel problem-solving and iterative refinement, suggest that Anthropic's investments in model reasoning are translating into genuine capabilities beyond benchmark metrics. This positions Opus 5 as potentially the most practical large language model currently available.
