Anthropic Launches Opus 5, Promising Near-Fable Performance at Half the Price

Anthropic rolled out Opus 5 on Friday, the latest update to its long-standing heavyweight model, positioning it as a cost-effective alternative to the more powerful Fable 5. Priced at $5 per million input tokens and $25 per million output tokens — half the cost of Fable 5 and slightly cheaper than OpenAI’s GPT-5.6 — the model marks a deliberate shift toward token efficiency rather than a radical capability leap, according to analysts.

“It comes close to the capabilities of Claude Fable 5 in many domains,” Anthropic said in a release, as reported by The Verge, and is particularly stronger at complex coding tasks. The company highlighted that Opus 5 wrote its own computer vision pipeline during benchmark testing in response to an incomplete prompt, according to a TechCrunch report.

**Performance and Pricing**

Benchmark results shared by Anthropic show Opus 5 performing at about the same level or slightly ahead of Fable 5 on coding tasks such as Frontier-Bench and DeepSWE, according to Ars Technica. It also beats Opus 4.8 and OpenAI’s GPT-5.6-Sol across most task categories. However, the Ars Technica analysis describes the performance gains as iterative rather than a breakthrough, with the main selling point being a model “just shy of Fable at approximately half the cost.”

The pricing remains unchanged from the predecessor Opus 4.8, which launched May 28. Anthropic is also introducing a “Fast mode” in research preview that doubles the speed for double the price, as reported by The Verge. Customers can opt into an Automatic Fallbacks beta, which routes requests to a less powerful model when a prompt triggers a safety classifier, ensuring a functional response instead of an error, according to TechCrunch.

**Lighter Safety Restrictions**

Unlike Fable 5 and Mythos 5, Opus 5 is not subject to the 30-day data retention policy that raised privacy concerns among some users, TechCrunch reported. Ars Technica confirmed the model does not carry the same controversial protections as Fable. Anthropic, however, said in its release that Opus 5 includes stronger cybersecurity safeguards than Opus 4.8 and is “the most aligned Opus model,” according to The Verge.

The safeguards are calibrated to be less restrictive than those on Fable 5. Anthropic expects classifiers to engage 85% less often for Opus 5, TechCrunch noted. The model blocks use for scanning vulnerabilities in compiled software binaries but permits source code analysis, since that is “more likely to be used for defensive purposes,” TechCrunch reported.

Still, Opus 5 lags behind in cybersecurity exploitation skills. Anthropic acknowledged that due to training decisions, the model is “substantially behind Mythos 5 on the exploitation of those vulnerabilities,” according to Ars Technica, which noted the company deliberately avoided cutting-edge cybersecurity training for Opus 5.

**Regulatory Context**

The launch follows a period of intensified government scrutiny of Anthropic’s models. The Fable 5 model — a public-facing variant of the Mythos class — drew the attention of the Trump administration, was briefly taken offline, and reintroduced with stronger cyber safeguards, as reported by The Verge. That episode kicked off a new regulatory era, which also saw OpenAI release GPT-5.6 in a limited preview restricted to government-approved entities.

When asked whether the Trump administration tested Opus 5 before public rollout, Anthropic spokesperson Danielle Ghiglieri told The Verge, “We continue to work with our government partners to conduct their own independent testing of our models. This includes Opus 5.”

Anthropic is marketing the model primarily to enterprise customers for knowledge work and biology, positioning it as an ideal everyday model while reserving Fable 5 for the most ambitious projects and long-horizon AI agent work, according to The Verge.

With only the lightweight Haiku model still awaiting a 5-series upgrade, Opus 5’s arrival completes the upper tier of Anthropic’s product line — now offering a clear price-performance trade-off in a market where both regulators and customers are watching costs closely.