Anthropic has launched Claude Opus 5, positioning the model as a practical, high-performance option for everyday coding, research, business automation, and scientific work, while keeping API pricing unchanged from its Opus 4.8 predecessor.
TL;DR
- Claude Opus 5 is available across Anthropic’s platforms and is the new default model for Claude Max.
- It costs $5 per million input tokens and $25 per million output tokens.
- Anthropic says it delivers state-of-the-art results across several coding and knowledge-work evaluations.
- Fast mode offers around 2.5 times the standard speed at twice the base price.
Anthropic is expanding its Claude 5 lineup with Claude Opus 5, a model designed to balance frontier-level intelligence with the cost controls businesses need for regular production use.
According to Anthropic’s official announcement, Opus 5 comes close to the intelligence of Claude Fable 5 at half the price, while becoming the company’s new default model on Claude Max and its strongest model on Claude Pro. The company is charging $5 per million input tokens and $25 per million output tokens, matching the pricing of Opus 4.8.
TechCrunch reported that Opus 5 arrived only two months after Opus 4.8 and follows Anthropic’s June releases of Mythos 5, Fable 5, and Sonnet 5. The publication noted that Opus 5 could be preferable for many use cases because it is cheaper and less restrictive than Fable 5, while outperforming it on some announced benchmarks.
Anthropic says the model is “much stronger at verifying its work and iterating carefully until it succeeds.” That focus is reflected in its coding, computer-use, automation, and research evaluations.
On Frontier-Bench v0.1, Anthropic said Opus 5 more than doubled Opus 4.8’s performance at a lower cost per task. On ARC-AGI 3, its score was three times higher than the next-best model, while its pass rate on Zapier AutomationBench was around 1.5 times the next-best result at the same cost.
The company also reported improvements in scientific work. Opus 5 scored 10.2 percentage points higher than Opus 4.8 on an internal organic chemistry benchmark and gained 7.7 percentage points on a protein-variation task.
“Claude Opus 5 topped Zapier’s AutomationBench leaderboard without spending more tokens than prior Claude models,” said Zapier CEO Wade Foster. Anthropic said the model completed an end-to-end churn-prevention workflow that previous models failed to finish.
CNET characterized Opus 5 as Anthropic’s new everyday assistant, highlighting the broader push to make advanced coding, knowledge-work, and autonomous task capabilities useful in regular workflows rather than reserving them only for specialized frontier-model use cases.
Safety remains part of the launch. Anthropic said Opus 5 recorded a 2.3 score for overall misaligned behavior in its automated audit, its lowest among recent models. Its cyber classifiers are expected to intervene around 85% less often than those used for Fable 5, although requests involving exploit generation, penetration testing, and binary vulnerability scanning remain restricted.
Topics for more insights:
Anthropic is also introducing automatic fallbacks in beta, allowing flagged API requests to route to another model instead of ending with an error. Developers can also change available tools during a conversation without invalidating the prompt cache.
Together, the changes position Opus 5 as a more economical production model for developers and enterprises that want stronger agentic performance without paying for Anthropic’s most advanced frontier system.
