Artificial Analysis puts Opus 5.5 at $5.98 per task, near Opus 5
The evaluator said a 20% price cut and cheaper cache reads offset a token-usage increase that would otherwise have pushed the figure to $10.51.
Artificial Analysis said Claude Opus 5.5 costs $5.98 to run its Intelligence Index, against $5.86 for Opus 5, both at maximum reasoning. The evaluator published the figures on Tuesday, hours after Anthropic released the model. It said the two are similar.
Getting there took two price cuts against one increase, Artificial Analysis said. Opus 5.5 uses more tokens than Opus 5, which on its own would have raised cost per task by about 80%, to $10.51. Anthropic's 20% cut to base prices brought that to $8.41.
Cutting cache reads to $0.20 per million tokens took a further 29% off, the evaluator said, landing at $5.98. The 29% is a further cut on top of the base-price reduction, not a restatement of it. PhotonCap put the cache-read cut at 60% against Opus 5, according to its summary of Anthropic's announcement.
On OpenAI's side, Artificial Analysis said GPT-6 Sol and Luna halve the cost of the models they replace while scoring about the same on its index. It put the driver at a roughly 50% reduction in token pricing rather than any change in how much the models write.
Simon Willison, who writes about language models, said GPT-6 Luna is half the price of GPT-5.6 Luna, which he had called cheap. These numbers may not hold for a particular workload, since the index is one evaluator's own mix of tasks.