The AI pricing war just got a lot more interesting. OpenAI’s GPT-5.6 Sol has arrived with a value proposition that reads like a Black Friday ad: half the input price and roughly 40% lower output costs compared to Anthropic’s Claude Fable 5, with benchmark performance that actually keeps pace.

Sol’s pricing comes in at $5 per million input tokens and $30 per million output tokens. Fable 5, by comparison, charges $10 per million input tokens and $50 per million output tokens. Quick napkin math: a workload consuming 1 million input tokens and 1 million output tokens costs $35 with Sol versus $60 with Fable.

That’s a 42% savings on a blended basis. Scale that across an organization processing billions of tokens monthly and you’re looking at serious budget relief.

In Terminal-Bench 2.1 testing, Sol scored 88.8% on agentic coding benchmarks. Fable 5 posted 84.3%. Sol didn’t just match the more expensive model on execution tasks. It beat it.

The token efficiency story is arguably even more impressive. Sol is reported to use 77% lower token costs per task, with 80% to 94% less code written compared to prior systems and Fable-based workflows.

Sol appears optimized for a specific slice of the AI workload spectrum: execution tasks, command-line operations, coding pipelines, and API-heavy workflows where speed and cost per token are king. Fable 5, on the other hand, reportedly maintains an edge in deep analysis and creative outputs.