Anthropic released Claude Opus 5 on July 24, 2026, a heavyweight model priced at $5 per million input and $25 per million output tokens — matching Opus 4.8's cost while closing on the frontier intelligence of Fable 5 at half the price. The launch caps a rapid summer for the company, following Mythos 5, Fable 5 and Sonnet 5 into the "5" family in under two months.
State of the art on frontier benchmarks
Opus 5 lands as the new leader on Frontier-Bench v0.1, more than doubling Opus 4.8's score at a lower cost per task. On CursorBench 3.2 it sits within 0.5% of Fable 5's peak at half the price, and on ARC-AGI 3 it scores three times higher than the next-best model. OSWorld 2.0 and Zapier AutomationBench also highlight the model's efficiency gains for agentic and computer-use workloads. Anthropic said the model is now the default on Claude Max and the strongest option on Claude Pro.
Agency, safety and cyber safeguards
The company positions Opus 5 as its most aligned model to date, citing the lowest scores on its automated misalignment behavioral audit and stronger adherence to Claude's Constitution than any prior release. Cybersecurity safeguards are tuned to intervene about 85% less often than on Fable 5, blocking binary vulnerability scanning and exploit generation while permitting source-code vulnerability research. A new Automatic Fallbacks beta reroutes flagged API requests to less-restricted models rather than returning errors.
Pricing, Fast mode and developer availability
Opus 5 is available across all Anthropic platforms and through Claude Code, Claude Cowork and third-party IDEs including Cursor, JetBrains and Kiro. A Fast mode option runs the model at roughly 2.5x standard speed for twice the base token price. Anthropic emphasized that early-access customers — including Cognition, Cursor, Zapier, Latch Bio, Lovable and Box — reported large gains on long-horizon coding, deep research, life-sciences workflows and enterprise document analysis, often at a fraction of the token budget used by Opus 4.8.
The Opus 5 debut extends Anthropic's push to standardize its top-tier line on the Mythos-class architecture pioneered with Fable 5. It also signals that competition on cost-per-intelligence between OpenAI's GPT-5.6 family and Google's Gemini 3.6 Flash is intensifying, with pricing rather than raw capability increasingly deciding which foundation model developers adopt for production workloads.
Related coverage: Anthropic Ships Claude Sonnet 5, Its Most Agentic Midsize Model, Spirit AI Raises 1.5B Yuan Series A+ for Embodied Robot Brains, and MACHINA Physical AI Summit Debuts in Paris.
Reporting based on coverage from Anthropic, TechCrunch and SiliconANGLE.