Anthropic and OpenAI both announced new models on Sept. 22, putting Claude Opus 5.5, GPT-6 Sol and GPT-6 Luna into the same news cycle. The releases are not a clean head-to-head test, but their pricing and product pitches show where the companies are competing: lower operating costs for people building software with models.
Anthropic says Claude Opus 5.5 is available in Claude, Claude Code and the Claude Platform. Its listed API price is $4 per million input tokens and $20 per million output tokens, down from $5 and $25 for Opus 5. Anthropic also lists cache reads at $0.20 per million tokens, versus $0.50 previously. The company says its default-setting tests show a 40% reduction in typical workload costs and output generation more than 30% faster; those are Anthropic's own measurements.
Different jobs inside the stack
OpenAI's announcement puts GPT-6 Sol in the more demanding coding and agent-work lane. It lists Sol at $2 per million input tokens and $10 per million output tokens. GPT-6 Luna is the cheaper option for more focused, high-volume work, listed at $0.10 for input and $0.50 for output per million tokens. OpenAI says both are available through its API and in specified Work and Codex plans, with Chat availability rolling out separately.
The companies are also emphasizing different controls around their launches. Anthropic says Opus 5.5 is being deployed with safeguards similar to those used for its more capable models, including verification programs for biology and upcoming cybersecurity access. OpenAI presents Sol and Luna as additions to its GPT-6 family and describes its own factuality, coding and alignment evaluations.
That is why the price figures alone do not settle which model is better for a particular job. Token prices, caching, effort settings, tool use and the task itself all affect the bill and the result. Both announcements contain company-run comparisons, so developers choosing among them will still need to test the models against their own workloads.
Tech