Anthropic Launches Opus 5, a New Artificial Intelligence Model Featuring Lower Pricing and Fewer Restrictions
Anthropic releases its new flagship model, Opus 5, closely following the previous version, offering performance that surpasses Fable 5 in many aspects while also being much more privacy-friendly.
Stock photo for illustration only, not from the actual event
- Anthropic launched the Opus 5 model on Friday, July 24, 2026
- It is smaller than Fable 5 yet achieves higher benchmark test scores
- Exempt from the 30-day data retention policy and reduces safety filter activity by 85%
- Adds the Automatic Fallbacks feature to automatically route requests to another model when safety conditions are triggered
Anthropic has officially launched Opus 5, its latest top-tier artificial intelligence model, on Friday, July 24, 2026. It features a more compact size than the Fable 5 model, along with a lower usage price and fewer usage restrictions, making it better suited for a wider variety of applications than before.
This release comes after the Opus 4.8 model was recently released on May 28, 2026, just two months prior. The 5 series has been gradually rolling out various sub-models since June, including Mythos 5, Fable 5, and Sonnet 5, leaving only the smaller Haiku model awaiting its upgrade to the 5 series.
In its launch announcement, the company emphasized that Opus 5 demonstrates significantly enhanced capabilities in verifying its own work and iterating to fix issues until successful. For instance, it has the ability to write computer vision scripts on its own in response to incomplete initial prompts.
Additionally, Opus 5 breaks free from several constraints that previously posed obstacles for the Fable model since its launch. The model is no longer forced to operate under the 30-day data retention policy, which had previously caused concerns for privacy-focused user groups.
"much stronger at verifying its work and iterating carefully until it succeeds"
Despite greater flexibility, essential safety systems remain in place, particularly regarding cybersecurity such as creating scopes for hacking or penetration testing. The filters do not allow Opus 5 to scan for vulnerabilities in software binaries, but it is still permitted to search for vulnerabilities in source code due to the higher likelihood of defensive use.
Anthropic's release of the 5 series reflects a very fast-paced product development strategy to capture an advantage in the highly competitive artificial intelligence market. The fact that Opus 5 outperforms its predecessor, Fable 5, highlights how the "smaller yet smarter" architectural tuning approach is becoming a major trend that significantly reduces computing costs for developers.
To mitigate the impact of safety systems potentially interrupting workflows, Anthropic has developed a new tool called Automatic Fallbacks, available for users to enable in beta. This system automatically redirects requests to a less powerful model whenever a prompt triggers a safety filter, ensuring that API users receive actual working results instead of encountering error messages.
Source: TechCrunch
Found something wrong in this article? Report an issue with this article
Comments
Leave a Comment