OpenAI released GPT-6 Sol and GPT-6 Luna on Tuesday at roughly half the per-token price of the models they replace, about 90 minutes after Anthropic launched Claude Opus 5.5 with cuts of its own. Sol, aimed at coding and complex work, costs $2 per million input tokens and $10 per million output tokens, down from $4 and $20 for GPT-5.6 Sol. Luna, for high-volume clerical work, costs $0.10 and $0.50. OpenAI says the prices are permanent, not introductory.
Anthropic priced Opus 5.5 at $4 per million input tokens and $20 per million output, which it says is 20 percent cheaper per token than Opus 5 and around 40 percent cheaper to run on typical workloads. The company calls it the strongest model it has tested and says it beats its larger Fable model on many benchmarks. OpenAI says Sol makes about half as many mistakes as its predecessor.
The timing turned two launches into one price war. On list price alone, Sol now costs half as much as Opus 5.5, and both companies pitched their releases around what a finished task costs rather than raw capability. That matters most to businesses running agents that call a model thousands of times a day, where a halved token price compounds quickly, and it lands as startups such as Harvey are already shifting work to cheaper open-weight models to protect their margins.
Neither company published the numbers that would settle the comparison. OpenAI said Sol outperforms Anthropic’s models without quantifying by how much, and Anthropic’s own benchmark claims were similarly general. Nor did either say whether the cuts reflect cheaper compute or margin they have chosen to give up. A permanent price is only as permanent as the next launch.
