Anthropic Releases Opus 5.5 With Lower Prices and Fable-Level Performance
Anthropic Opus 5.5 has officially released as the company’s newest flagship AI model, setting what the team describes as a new state-of-the-art benchmark for coding and knowledge work performance, while simultaneously cutting prices significantly compared to its predecessor.
Opus sits at the top of Anthropic’s three-tier Claude lineup, positioned as the most capable and expensive option, with Sonnet occupying the middle tier and Haiku serving as the fastest, cheapest option. Notably, Anthropic says Anthropic Opus 5.5 outperforms the larger Fable model across many benchmarks and successfully completed a number of informal tasks that Fable was unable to finish.
A Significant Price Drop Alongside Better Performance
The new model isn’t just more capable, it’s also considerably cheaper to run. Output tokens for Anthropic Opus 5.5 are now priced at $20 per million tokens, down from $25 for the previous version, with similar price reductions applied across other usage metrics as well. The model also runs faster overall, reflecting a genuine drop in the computing resources required to serve it at scale.
Beyond raw performance and pricing, Opus 5.5 introduces meaningful changes to how the model actually communicates. According to Anthropic, the updated version relies less on technical jargon and more consistently places important information at the beginning of its responses, making outputs easier to scan quickly.
Part of a Rapid Release Cycle
This launch arrives just two months after Opus 5 debuted on July 24, reflecting Anthropic’s continued rapid iteration pace within the Claude lineup. According to the company’s announcement, Sonnet 5.5 and Haiku 5.5, the next tiers in the lineup, are expected to follow in the coming weeks with similar performance improvements applied across those models as well.
Safety Safeguards Tied to Biology and Cybersecurity
Anthropic says safety capabilities in biology and cybersecurity for Anthropic Opus 5.5 are comparable to its Mythos model, meaning its release falls under the same safeguards applied to the company’s Fable model. Those safeguards specifically limit how the models can be used to discover exploits within compiled programs or assist in developing recognizable biological weapons, among other restricted use cases.
Notably, Opus 5.5 marks Anthropic’s first model release since CEO Dario Amodei publicly embraced calls to deliberately pace AI capability development, aligning the pace of new releases more closely with the pace of alignment and safety research.
“I have become convinced that fully addressing the risks requires even more prudence,” Amodei wrote in a post published earlier this month, “not just investing in risk prevention, but pacing the rate of capabilities advancement so that risk prevention has time to keep up.”
How Safety Testing for Opus 5.5 Actually Worked
Opus 5.5’s safety training process remained broadly consistent with previous model releases, including alignment testing and pre-release evaluation conducted by outside organizations like METR and Frontier Design. However, Anthropic emphasized that more advanced training and evaluation systems are already being prepared for future model releases, including improved security and monitoring infrastructure designed to catch issues earlier in the development process.
“As AI becomes more capable, public policy should play a larger role in making sure the systems people rely on are safe,” Anthropic wrote in its announcement. “That capacity takes time to build, and we’ve started to put the infrastructure in place to support it. We expect to share more details on these efforts soon.”
What This Signals Going Forward
Opus 5.5’s release, arriving so soon after Amodei’s public commitment to pacing AI development more deliberately, suggests Anthropic is attempting to demonstrate that meaningful capability and pricing improvements can still happen on a reasonably fast timeline, even while the company publicly commits to weighing safety considerations more heavily than raw speed of development going forward. Whether that balance holds as Sonnet 5.5 and Haiku 5.5 roll out in the coming weeks will likely offer an early test of how sustainable this approach actually proves to be.