OpenAI has launched three new models under GPT-5.6, challenging Anthropic's Fable 5 with competitive pricing and performance.
OpenAI has unveiled three distinct models under the GPT-5.6 series—Sol, Terra, and Luna—marking a significant advancement in AI technology. The primary focus is on comparing Sol with Anthropic's Fable 5, its chief competitor.
Pricing reveals a stark contrast: Sol charges $5 per million input tokens and $30 for outputs, while Fable 5 is priced at $10 for inputs and $50 for outputs, effectively making it double the cost of Sol. In a notable twist, Luna, the most economical of the three, costs just $1 for input and $6 for output, and has already outperformed Anthropic's Opus 4.8 in coding tasks.
Fable 5 has encountered challenges following a U.S. government ban due to security vulnerabilities, leading to a 19-day suspension. Since re-launching on July 1 with limited access, Anthropic has been under pressure to clarify its pricing structure, delaying the implementation of a usage-credits paywall multiple times. The latest deadline is set for July 19, raising concerns that if Fable is removed from subscriptions, Anthropic's offering would revert to a less competitive model.
We are extending access to Claude Fable 5 for all paid plans, while also increasing Claude Code’s weekly rate limits by 50% until July 19.
— Claude (@claudeai) July 12, 2026
Benchmark comparisons reveal a competitive landscape: Sol scored 80 on the Artificial Analysis Coding Agent Index, while Fable 5 achieved 77.2, using half the tokens and completing tasks in less than half the time. In broader evaluations, Fable 5 narrowly edges out GPT-5.6 by a single point on the Intelligence Index, suggesting minimal differences in capabilities.
Comparative Model Testing
Testing has primarily focused on coding abilities, but a more creative prompt assessed both models' narrative skills. In a time-travel scenario, both produced novella-like outputs, but neither recognized the inherent paradox. Sol's narrative, "The First Fire," features a straightforward sci-fi approach, while Fable’s "Lo Que Arde, Vuelve" offers a culturally rich tale, although both narratives could improve in execution.
In terms of associative thinking, Sol's metaphorical connections on worker exploitation felt disjointed compared to Fable’s more seamless integration of themes. Despite Fable's strengths, it occasionally indulged in overly complex language. Overall, preferences between the two models may vary based on clarity versus subtlety.
In a logical reasoning test, both models misinterpreted the prompt's constraints, leading to the same incorrect conclusion of 17 minutes for a bridge puzzle. However, Fable provided a clearer rationale for its answer. In a coding challenge, Fable significantly outperformed Sol, showcasing a more engaging and polished user experience.
Conclusion
While both models have their strengths and weaknesses, Fable 5 appears to be the more versatile option, especially for non-technical users. However, the pricing differences could heavily influence user decisions, particularly as Fable 5 faces uncertainty in its subscription model moving forward. As the AI landscape evolves, the choice between these models will largely depend on user needs and budget considerations.
