Anthropic Launches Claude Opus 5, Undercutting Its Own Flagship at Half the Price

Anthropic released Claude Opus 5 on July 24, positioning the new model as a faster, dramatically cheaper alternative to its top-end Claude Fable 5 — and, on several published benchmarks, a better one. The launch, covered by TechCrunch, Fortune and Axios, marks the company's fourth Claude 5-family release in less than two months, an unprecedented cadence for a lab that once shipped flagship models annually.

Opus 5 is priced at $5 per million input tokens and $25 per million output tokens — unchanged from its predecessor Opus 4.8, and half the cost of Fable 5's $10 and $50 rates. Fortune reported that the model delivers performance close to Fable on many tasks despite being a smaller model, and Anthropic's own announcement shows it outperforming Fable 5 outright on a number of benchmarks.

A dial for cost versus capability

The headline feature is a user-controllable effort toggle. Developers and Claude app users can set how much work the model expends on a task — low, medium or high — trading cost and latency against capability on a per-request basis. The model ships with a 1 million token context window, 128,000-token maximum output, and thinking enabled by default.

Anthropic said Opus 5 is "much stronger at verifying its work and iterating carefully until it succeeds," citing internal demonstrations that included writing working computer vision pipelines from incomplete prompts. The model is now the default on Claude Max plans and the strongest model available to Claude Pro subscribers.

Fewer guardrails, new fallbacks

TechCrunch noted two quieter but consequential changes. Opus 5 ships with lighter safety restrictions than Fable 5 — Anthropic says safeguards engage roughly 85 percent less often — and it is not subject to the 30-day data retention policy applied to the Fable and Mythos tiers. Alongside the launch, Anthropic introduced an "Automatic Fallbacks" beta that routes a request to a less powerful model when safety classifiers trigger, rather than refusing outright.

The combination of lower price, lighter restrictions and near-flagship performance led several outlets to conclude the new model will be preferable to Fable 5 for most workloads — an unusual case of a lab aggressively undercutting its own premium product.

Pressure on every side

The release lands in a crowded window. OpenAI has been iterating rapidly on its GPT-5.x line, Chinese open-weight models continue to gain share among US developers, and enterprise buyers are increasingly price-sensitive as agentic workloads multiply token consumption. For Anthropic — which reported a $30 billion revenue run rate earlier this year — Opus 5 looks like a deliberate move to defend the high-volume middle of the market while Fable holds the frontier.

Opus 5 is available immediately across the Claude API, claude.ai, and major cloud platforms.