OpenAI, Anthropic slash prices as GPT-6 and Opus 5.5 ship

Anthropic released Claude Opus 5.5 on the same day OpenAI shipped GPT-6 Sol and GPT-6 Luna, one day after Grok 4.7 and MiMo v2.6 Flash/Pro landed. Writing on his blog, Simon Willison walks through what looks like a compressed release calendar and a new low water mark on price.

GPT-6 Luna is priced at $0.10 per million input tokens and $0.50 per million output, roughly half the going rate of GPT-5.6 Luna at $0.20/$1.20. "At $0.10/$0.50 GPT-6 Luna is one of the cheapest models OpenAI have ever released," Willison writes. GPT-6 Sol lands at $2 input and $10 output, which now matches the undercut Grok 4.7 had staked out at $2/$6.

Opus 5.5 answered with cuts of its own. Input drops to $4 per million and output to $20, about 20% below the prior $5/$25, with cached input reads falling 60% to $0.20 per million. Anthropic's Thariq Shihipar framed it as customer-driven, writing that "Opus 5.5 is the result of your feedback. It communicates clearly, it's cheaper per token" than the previous generation.

The launch was not clean. Running Willison's pelican-on-a-bicycle SVG test at Opus 5.5's "max" reasoning setting, the model burned through its 128,000-token output ceiling and returned nothing. "Those two failures each cost me $2.56 and took nearly 20 minutes," he writes, concluding that max reasoning is "effectively useless" for now. Two researchers on our radar circulated the post the same day.

Sonnet 5.5 and Haiku 5.5 are on Anthropic's near-term roadmap. "It's hard to overstate how competitive this pricing is," Willison writes.