Anthropic launches Claude Opus 5.5 with 20% list-price cut and ~30% speed gain; OpenAI ships GPT-6 Sol and Luna same day at 50% lower prices than GPT-5.6

Anthropic released Claude Opus 5.5, the first model in its 5.5 family and now the default in Claude Code and the Claude app. List price dropped from $5/$25 to $4/$20 per million input/output tokens, a 20% cut, not the 40% headline figure, which only holds at default effort once higher token usage is netted out. Anthropic claims roughly 30% faster and 40% cheaper per task than Opus 5 at default settings, though independent testing by Artificial Analysis found near cost parity with Opus 5 at maximum effort ($5.98 vs $5.86). The same day, OpenAI released GPT-6 Sol and Luna at roughly 50% below GPT-5.6 pricing.

placeholder

The "40-50% cheaper" framing from both companies is marketing arithmetic that does not hold at higher effort settings, where Opus 5.5 is roughly cost-neutral against its predecessor once real usage patterns are accounted for. Claims that Opus 5.5 matches or beats GPT-6 Astra on "most benchmarks" and fixes top user complaints about writing quality are self-reported by Anthropic and not independently verified in this extraction. No evidence links were captured, so the benchmark and safety claims cannot currently be checked against primary sources, and the source itself (Latent Space/AINews) is a newsletter aggregator rather than original testing.

1. 01Independent benchmark results (Artificial Analysis, Vals AI) comparing Opus 5.5 to Opus 5 and GPT-6 Astra at matched effort settings, not just Anthropic's own comparisons
2. 02Actual API spend data from teams running Claude Code or agent workloads at scale over the next 4-6 weeks, to see whether real bills fall by anywhere near 40%
3. 03Whether OpenAI or Anthropic make further price moves within 30 days, which would confirm this is a sustained price war rather than a one-off coincidence