OpenAI Squeezed At The Frontier-End By Anthropic’s Claude Fable 5.2, While Meta’s Muse Ember Gives Its Bargain-Class GPT-6 Luna A Bruise
By Google News AI ·
Today is not a good day for OpenAI, especially as it is getting hammered from two sides, with Anthropic's Claude Fable 5.2 leaving GPT-6 Astra in the proverbial dust, and Meta's Muse Ember squeezing GPT-6 Luna hard in the bargain category of AI models.
We noted towards the middle of October that Anthropic's Fable 5.5 AI model had been spotted in the wild. Well, the high-flying AI lab has just chosen to launch a different model, the Claude Fable 5.2.
According to Anthropic, "in early testing on FrontierCode 2.0, Fable 5.2 at High effort scored about level with Opus 5.5 at Max effort, for roughly half the cost per task. At Xhigh effort, it scored the highest of any released model tested."
What's more, according to Vals AI, at the "High" reasoning level, Claude Fable 5.2 costs only $17.94 per test, which is cheaper than both Opus 5.5 and GPT-6 Astra at a comparable reasoning effort.
Moreover, as per the preliminary benchmarks (in the image above), Claude Fable 5.2 outperforms Fable 5.1, Opus 5.5, OpenAI's GPT-6 Astra, and Google's Gemini 4 Carbon in nearly every category.
Anthropic also highlighted that Fable 5.2 agents were able to solve Fermat's Last Theorem in around 30 hours:
"A team of 1,000 Fable 5.2 agents fully formalized Fermat’s Last Theorem in 30 hours, translating the proof into a language a computer can check step by step."
"With the same setup, Opus 5.5 completed only 26% in 18 hours, and barely made any more progress after 30 hours."
Critically, Anthropic has priced its best model at $8 per 1 million input tokens and $40 per 1 million output tokens, down from $10 and $50, respectively, for Fable 5.1. Do note that Fable 5.2 is also around 30 percent faster than Fable 5.1.
Finally, do note that Anthropic is now giving users between 1 and 3 free usage resets, depending on how long they've been subscribed. Also, the Pro plans are now eligible for a 50 percent weekly usage allowance for Fable 5.2.
At the other end of the spectrum, Meta has just launched the very capable, albeit tiny, Muse Ember. The model outcompetes Claude Haiku 5.5 and GPT-6 Luna in nearly every category of benchmarks.
Muse Ember also scores 36 on the Artificial Analysis Intelligence Index, matching the score of Claude Haiku 5.5 (max) at a third of the cost per task ($0.13). Moreover, the model matches GPT-6 Luna’s API pricing - $0.10 per 1 million tokens of input and $0.50 per 1 million tokens of output.
About the author: Rohail Saleem reports on the business of AI and semiconductors for Wccftech: the chips, data centers and compute deals behind companies such as NVIDIA, AMD, TSMC, OpenAI and Anthropic, and what those deals mean for the hardware people buy.
He has written more than 2200 articles for Wccftech since 2019, often reading earnings reports, supply-chain data and analyst notes to explain where the money and the hardware are going. He holds a commerce degree from the University of Toronto's Rotman Commerce program.
Follow Wccftech on Google to get more of our news coverage in your feeds.
Originally published by Google News AI
