Claude Opus 5.5 and GPT-6 Sol: comparing cost per task

On September 22, Anthropic released Claude Opus 5.5. According to a DEV Community article published on October 3, the release brought a fifth off the list price and 60% off cached reads. That marked the first Opus price cut since 4.5. About ninety minutes later, OpenAI released GPT-6 Sol and Luna, pricing both at half the cost of the models they replace. The two announcements landed on the same day, putting the two companies' new models in direct comparison.

The DEV Community article, however, argues that for anyone paying per token, the Opus 5.5 pricing is only half the story. The other half is how many tokens each model spends to finish a task. Token price alone does not determine the total cost of completing a task; the number of tokens consumed matters just as much. The article says independent measurements of this token consumption did not match the claims made at launch. It does not specify which model comes out cheaper per task, nor does it provide the task types or token counts behind those independent measurements.

So far, the public record shows a clear sequence: Anthropic moved first with a price reduction on Claude Opus 5.5, OpenAI followed roughly an hour and a half later with GPT-6 Sol and Luna at half the price of their predecessors, and the discussion has since turned to per-task cost. The article stops short of declaring a winner. It leaves the comparison at the point where list price and token consumption need to be weighed together, with the independent measurements and the launch claims still in disagreement.