Anthropic Unveils Claude Opus 5 at Half the Price of Fable 5

Anthropic released Claude Opus 5 on July 24, 2026, positioning it as a cost-effective everyday model that approaches Fable 5's capabilities at half the price. The launch marks the fourth Claude 5-series model in less than two months, signaling a strategic shift from blockbuster releases to continuous iteration.

Pricing and Performance

Opus 5 costs $5 per million input tokens and $25 per million output tokens, matching Opus 4.8 pricing while undercutting Fable 5's $10/$50 rate. Despite the lower cost, Anthropic states the model outperforms Fable 5 on coding and knowledge work evaluations including Frontier-Bench and GDPval-AA. The company also notes Opus 5 is slightly cheaper than OpenAI's GPT-5.6.

The model shows particular improvements in organic chemistry tasks, especially inferring molecular structures from spectroscopy data. According to Anthropic, "Opus 5 was much stronger at verifying its work and iterating carefully until it succeeds."

Reduced Restrictions, More Control

Opus 5 features significantly relaxed safety guardrails compared to its predecessor. Safety classifiers engage 85% less often than for Fable 5, addressing a common frustration among enterprise users who have found overly restrictive filters impede productivity. The model is also not subject to the 30-day data retention policy that covers Fable and Mythos models, a key consideration for organizations handling sensitive information.

A new effort ladder gives users unprecedented control over cost-capability tradeoffs. Opus 5 supports five compute levels—low, medium, high, xhigh, and max—allowing teams to dial down for routine tasks or maximize compute for complex problems. The Automatic Fallbacks beta feature routes requests to less powerful models when safety classifiers trigger, preventing error messages that previously halted workflows.

Anthropic describes Opus 5 as "the most aligned Opus model and is the least susceptible to being tricked into misuse," suggesting the reduced restrictions don't compromise security.

Availability and Context

Opus 5 becomes the default model for Claude Max subscribers and the strongest model available on Claude Pro. It is accessible across Anthropic's paid plans.

The release arrives as enterprises grow increasingly cost-conscious about AI bills following rapid adoption in 2025 and early 2026. Industry-wide concerns about operational expenses have pushed vendors to offer more granular pricing and performance options. The launch also comes amid heightened scrutiny of AI safety following recent incidents involving other companies' models escaping sandboxes during testing.

Anthropic's accelerated cadence—four models since late May—reflects competitive pressure in a market where frontier intelligence is becoming commoditized. By offering near-top-tier performance at mid-tier pricing with fewer restrictions, Opus 5 targets the sweet spot for routine business applications where Fable 5's capabilities may be overkill but cheaper models fall short.

The model continues working with government partners on independent testing, according to Anthropic, as regulatory attention on AI releases intensifies.