OpenAI's Sol closes the gap on Anthropic at one-third the price
A one-point benchmark gap and a two-thirds price advantage mean Sol will drive enterprise adoption fast, reshaping which models surface your content in AI answers.
Key takeaways
- GPT-5.6 Sol scores 59 on the Artificial Analysis Intelligence Index, one point behind Claude Fable 5, at one-third the cost.
- Sol outperforms every competitor in agentic coding, the capability enterprise technology teams now evaluate most seriously.
- A 3x cost advantage is a strong adoption signal: the model enterprises choose determines which content gets cited in AI-generated answers.
- Institutions with tight budgets, including multilaterals and policy bodies, face a materially different build-versus-buy calculation with Sol available.
- OpenAI's simultaneous dominance in consumer AI search and enterprise infrastructure concentrates citation power in one provider.
One point. That is the margin separating OpenAI's GPT-5.6 Sol from Anthropic's Claude Fable 5 on the Artificial Analysis Intelligence Index, according to The Decoder. Sol scores 59; Fable 5 scores 60. The price gap is rather less delicate: Sol costs $1.04 per task, roughly one-third of what Anthropic charges for its flagship model.
That arithmetic is the story. When frontier capability compresses to a one-point benchmark gap while cost compresses by two-thirds, enterprise buyers face a straightforward procurement question. The answer, increasingly, will not favour the premium-priced incumbent.
Where Sol actually wins
The benchmark parity is striking, but the agentic coding result is the more consequential finding. The Decoder reports that Sol outperforms every competitor in that category. Agentic coding, where a model plans, writes, tests, and iterates across long-horizon tasks, is the capability that enterprise technology teams now evaluate most seriously when choosing AI infrastructure. It is also the domain most directly connected to productivity gains that CFOs can quantify.
For Anthropic, which has built a substantial portion of its enterprise positioning around Claude's performance on complex reasoning and developer workflows, Sol's lead in agentic coding is a more direct competitive threat than any summary benchmark score.
What model compression means for brands that depend on AI systems
Senior marketers and communications leaders at large institutions may regard a model pricing war as something that belongs to their technology colleagues. That is a mistake.
The models that enterprises deploy at scale determine which content gets processed, synthesised, and surfaced in AI-generated answers. Financial services firms, multilateral institutions, and large industrial groups are increasingly building internal AI tools on top of frontier models. The model they choose shapes what their employees find when they ask an AI system a question, and, critically, what the model cites when constructing that answer.
If cost drives enterprise adoption toward Sol, the citation and synthesis patterns that emerge from those deployments will reflect Sol's training, retrieval preferences, and content-weighting behaviours. Brands and institutions that have invested in optimising their content for AI visibility need to understand which models their target audiences are actually using. A 3x cost advantage is a powerful adoption signal.
For organisations like multilateral institutions and policy bodies, where budget constraints shape technology choices as surely as they do in the private sector, a frontier-quality model at one-third the cost changes the build-versus-buy calculation entirely. UNDRR or CGAP running AI-assisted research tools on Sol rather than Fable 5 is a plausible near-term outcome. The content those tools surface and trust will matter enormously to any organisation trying to maintain authority in AI-mediated information flows.