Anthropic Is Testing Fable 5.2 and Opus 5.5 Simultaneously, With Opus 5.5 Reportedly Landing This Week

Summary

AI model tracking community testingcatalog.com and several related outlets report this week that Anthropic has two unreleased Claude models in stealth testing simultaneously: Fable 5.2 and Opus 5.5. Fable 5.2 has been observed in internal testing and reportedly improves on Fable 5.1’s reasoning and creative performance. The more significant near-term news is Opus 5.5, identified in community testing under the internal codename claude-wafer-eap. Reports suggest it outperforms the current Claude Opus 5 on coding and long-running agent tasks and is being positioned to challenge GPT-6, with leaked pricing indicating it would cost less per token than the current Opus 5. Multiple trackers are pointing to a Tuesday, September 22 release for Opus 5.5. Anthropic has made no official announcement.

Why it matters

If Opus 5.5 launches this week at a lower price with better performance, it is a meaningful upgrade for the millions of people who pay for Claude Pro or use Claude through third-party apps. A better model at a lower cost is a straightforward win for everyday users. Watch Anthropic’s newsroom this week.

HTD Says

Anthropic may be dropping a new Claude model tomorrow, and if the leak on pricing holds, it is a faster and cheaper version of Opus 5, which is saying something. It is impossible to keep up with all the models from different providers and what each one brings in terms of improvements, intelligence, and functionality.

If a model is more efficient, it can perform similar tasks at a lower cost. In my experience, this isn’t always true. Some models are truly steps backward, even though they are newer. I’m currently on Sonnet 5 Medium (as of this writing), which works well for me. And I’m not going to buy credits to use something newer when previous models work fine. But that is until they actually remove that model from the interface.

Source:

testingcatalog.com