Claude Opus 5 takes the crown and costs half what Fable 5 does
Anthropic released Claude Opus 5 on July 24, and within hours it climbed to the top spot on the Artificial Analysis Intelligence Index. It sits above Claude Fable 5, GPT-5.6 Sol, and Kimi K3. That alone would be a story. But the pricing is what makes it interesting.
Opus 5 costs $5 per million input tokens and $25 per million output tokens. The same as Opus 4.8. Fable 5, which Opus 5 now beats on several benchmarks, costs roughly double. So Anthropic just shipped a model that outperforms its own flagship on key metrics at half the price.
On Frontier-Bench v0.1, Opus 5 more than doubles Opus 4.8’s performance at a lower cost per task. On CursorBench 3.2 at max effort, it comes within 0.5% of Fable 5’s peak score but at half the cost. On ARC-AGI 3, its score is three times the next-best model. On Zapier AutomationBench, its pass rate is roughly 1.5x the runner-up at the same cost per task.
The pattern across benchmarks is consistent. Opus 5 delivers near-Fable intelligence at Opus-tier pricing.
Anthropic positions Opus 5 as a model designed for daily use rather than occasional heavy lifting. It is the new default on Claude Max and the strongest option on Claude Pro. The effort-setting system lets developers trade intelligence for speed and cost, which matters for production workloads where you do not always need maximum reasoning depth.
The most striking anecdote from the release involves a Frontier-Bench task where the model received a drawing of a machine part and was asked to reconstruct it as a 3D FreeCAD model. The catch: the model was given no way to view the drawing directly. Opus 5 wrote its own computer vision pipeline to extract geometry from raw pixels, then rebuilt the full part. It succeeded repeatedly. No competing model could solve it after five attempts.
That kind of autonomous problem-solving is what Anthropic means when they call the model “thoughtful and proactive.” Multiple early-access partners reported similar behavior. Cursor saw Opus 5 come within a hair of Fable 5 on their internal benchmark. Zapier’s CEO said it ran a full churn-prevention workflow end to end without intervention. A trading firm engineer used it to build a market data feed for a new exchange in a single session, including writing its own test harness when no live feed was available for validation.
On safety, Opus 5 is notably less restrictive than Fable 5. Anthropic expects their cyber classifiers to trigger about 85% less often. The model can search for vulnerabilities in source code but is blocked from binary-based scanning, penetration testing, and exploit generation. It was not trained on cyber tasks, yet improved on them anyway through general capability gains. It approaches Mythos 5 at finding vulnerabilities but remains far behind at actually exploiting them.
This deliberate gap between detection and exploitation is the safety story. Anthropic is betting that a model strong enough to find bugs but deliberately hampered at weaponizing them will avoid the regulatory pressure that hit Fable 5 and Mythos 5 earlier this year.
The launch also includes two API features in beta. Mid-conversation tool changes let developers swap which tools Claude can use without busting the prompt cache. Automatic fallbacks route flagged requests to a less powerful model instead of returning an error, so API users always get a functional response.
Opus 5 arrives two months after Opus 4.8. The full Claude 5 family is now complete except for Haiku. That is a fast cadence, and it puts pressure on OpenAI and Google to respond at a similar tempo.
Simon Willison noted the buzz is positive and that the model sounds “relentlessly proactive” based on the release notes. The HN thread discussing Opus 5’s leaderboard position hit 320 points and 187 comments within 17 hours, which tells you the developer community is paying attention.
Opus 5 is available now on all platforms. Fast mode runs roughly 2.5x the default speed at double the base price.