Anthropic released Claude Opus 5.5 on September 22 with a 20% price cut, and OpenAI responded minutes later with two new GPT-6 variants, Sol and Luna, priced at half their predecessors' cost1. The back-to-back launches mark the sharpest direct pricing confrontation between the two labs to date[1].
Claude Opus 5.5 is priced at $4 per million input tokens and $20 per million output tokens. Anthropic said the new pricing works out to 40% less on a typical workload. The company positions Opus 5.5 as its main mass-market workhorse model, used for tasks like coding and other complex knowledge work2.
OpenAI's Sol and Luna are described as the latest versions of its middle-of-the-road or smaller models, focused on efficiency and speed. OpenAI priced both at half what their predecessors cost. The release came minutes after Anthropic's own announcement.
Simon Willison characterized the simultaneous drops as "a new price war"3. Ars Technica framed both releases under the same thesis: "a little more for a lot less money".
ANALYSIS The timing of OpenAI's release, minutes after Anthropic's, suggests each lab had visibility into the other's launch window and chose to compete on the same news cycle rather than cede attention. Both companies are compressing the cost curve from different positions in the model stack: Anthropic is cutting the price of its flagship reasoning tier, while OpenAI is halving costs on its efficiency-oriented models. The practical effect for API consumers is a rapid reduction in per-token spend across two of the three major frontier providers.
OpenAI had already been investing in cost-reduction infrastructure around GPT-6, disclosing improvements to prompt caching with higher hit rates, explicit breakpoints, and new diagnostics earlier the same day[3].