Pressure from two sides
According to Wccftech, OpenAI is being squeezed from two directions in the model market. At the frontier end, Anthropic's Claude Fable 5.2 outcompetes OpenAI's GPT-6 Astra in nearly every benchmark. At the budget end, Meta's lightweight Muse Ember model is positioned as a rival to GPT-6 Luna.
What Fable 5.2 beats
The report says Claude Fable 5.2 outperforms Fable 5.1, Opus 5.5, GPT-6 Astra and Google's Gemini 4 Carbon in comparative tests. Benchmarks are standardised tests used to compare how well models perform across different tasks.
Competition in the budget tier
Meta's Muse Ember is described as a lightweight model designed for agent tasks. Wccftech says it is putting pressure on OpenAI's more affordable GPT-6 Luna model.
An early assessment
The comparisons in the report come from a single outlet's account; independent test results and pricing details are not shared. How the models perform in real-world use therefore remains unclear.
