Anthropic releases Claude Opus 5 at half the price of Fable 5

Anthropic released Claude Opus 5 on 24 July 2026, positioning it as the default model for Claude Max subscribers and the strongest option available on Claude Pro. Anthropic said the model delivers close to Claude Fable 5-level intelligence at roughly half the cost.

The model is priced at $5 per million input tokens and $25 per million output tokens, the same as Opus 4.8 but about half what Fable 5 charges. It offers a one-million-token context window and up to 128,000 output tokens.

Claude Opus 5 is available through Claude, Claude Code, the Claude API under the model ID claude-opus-5, Amazon Bedrock, Google Cloud Vertex AI and Microsoft Foundry, according to SiliconANGLE.

On performance, the model outscored Fable 5 on 8 of 13 benchmarks. It scored 9.7 percentage points higher on Frontier-Bench, a set of 74 physics, chemistry and cryptography tasks, reaching 43.3%, and 8.5 points higher on AutomationBench.

Anthropic said Claude Opus 5 scored zero on its internal benchmark for exploiting vulnerabilities and can search code for security issues while blocking scanning methods commonly used by attackers. The company said its safety guardrails trigger 85% less often than they did on Fable 5, calling it its safest model to date.

The release adds two API features: the ability to interrupt a model mid-workflow and change its tools, and rerouting of prompts blocked by safety guardrails to a different model. In one test cited by Anthropic, the model verified its own output by writing "a computer vision pipeline to pull the geometry from the raw pixels".

Claude Opus 5 is Anthropic's fourth model release in under two months, following Mythos 5, Fable 5 and Sonnet 5 in June. A checkpoint codenamed "Claude Honeycomb EAP" briefly appeared in Cursor's model picker on 9 July ahead of the launch. Anthropic has not detailed further releases beyond Opus 5.