OpenAI responded to Opus 5.5 on time—but not up to par. It has just launched GPT-6 Sol at half the price of Sol 5.6, placing the model at the frontier of cost-per-intelligence. Artificial Analysis measured a 48 on the Intelligence Index at maximum effort, at approximately $1.06 per task. Opus 5.5 leads at 58. Astra and Fable 5.1 tie at 53. Opus 5 stands at 51. Grok 4.7 at 46. On API: $2 input and $10 output per million tokens, with caching at $0.20. Sol 5.6 was $4/$20. Astra and Fable charge $10/$50. Opus 5.5 is at $4/$20. In practice, Sol doesn’t compete for the ceiling—it competes for scale. On the Coding Agent Index, it rises to 57 and enters the cost-efficiency Pareto frontier. Intelligence remains at Sol 5.6 levels, but cost per task is halved. Hallucinations decrease. In GDPval and Briefcase, AA still observes regression. Opus 5.5 retains the Intelligence Index ceiling. Astra still uses fewer tokens at the top end. Fable remains Anthropic’s expensive option. Grok 4.7 still competes on sticker price. Sol becomes OpenAI’s affordable mid-tier model for high-volume work.