Claude Opus 5.5, GPT-6 Sol, GPT-6 Luna, and a new price war

Yesterday, Anthropic unveiled Claude Opus 5.5 and OpenAI released GPT-6 Sol and GPT-6 Luna. The new models have stirred up a price war in the LLM landscape. GPT-6 Luna is half the price of its predecessor, GPT-5.6 Luna, while GPT-6 Sol has a similar discount compared to GPT-5.6 Sol. GPT-5.6 Luna was previously Anthropic's top pick for building applications due to its excellent performance and affordability.

However, GPT-6 Luna now offers even better value at half the price, reducing any reasons to stick with GPT-5.6 Terra. GPT-6 Luna is the cheapest OpenAI model ever released, surpassing even GPT-4.1 Nano and GPT-5 Nano. Claude Opus 5.5 appears to address Opus's communication issues, being cheaper per token than Opus 5.0 with the intelligence of Fable 5.1.

It is also token-efficient and works across all effort levels. The price for Opus 5.5 has dropped by 20%, now at $4/million for input and $20/million for output. The new price is identical to GPT-5.6 Sol's price, but OpenAI has since halved their Sol prices. Other models like Sonnet 5.5, Haiku 5.5, GPT-6 Astra, and Claude Fable 5.1 are expected to follow soon.

Claude Opus 5.5 hit a maximum output token limit while reasoning about an SVG test, suggesting that the "max" level may not always yield the best results. Despite this, GPT-6 Sol and Claude Opus 5.5 have become the default models for Codex and Claude Code, with Anthropic's Datasette Agent demo upgraded to use GPT-6 Luna.