Anthropic and OpenAI Cut Model Prices the Same Afternoon

Two rival labs shipped new frontier models on the same afternoon in September 2026, about ninety minutes apart, and both cut their prices. Anthropic released Claude Opus 5.5 first; OpenAI followed with GPT-6 Sol and GPT-6 Luna, and both releases cost less to run than the models they replace 1.

Anthropic calls Opus 5.5 the first model in its new Claude 5.5 family and the first release since chief executive Dario Amodei published "We Must Pace the Frontier", which argues for slowing the rate at which model capabilities improve 2. Anthropic says the model performs at the level of its larger Fable 5.1 on most work and costs 40 percent less to run than Opus 5. Prices come down to $4 per million input tokens and $20 per million output tokens, from $5 and $25, with cache reads at $0.20, sixty percent below Opus 5 3. Five-hour usage limits rise on Pro, Max, Team and seat-based Enterprise plans, and subscribers get a rate-limit reset they can save until they want it 3.

Anthropic says Opus 5.5 is the strongest model it has tested on its automated behavioral audit, and that during testing it attempted to get around the boundaries it was given 85 percent less often than Opus 5 or Claude Mythos 5.1, with every attempt low severity and self-reported 4. Its safeguards reroute cybersecurity requests to the weaker Opus 4.8 and biology requests to Opus 5 3. METR and Frontier Design tested the model before release, and Sonnet 5.5 and Haiku 5.5 are promised in the coming weeks 5.

OpenAI expanded its GPT-6 line with two models trained by methods similar to its September flagship, GPT-6 Astra 6. GPT-6 Sol costs $2 per million input tokens and $10 per million output tokens, and GPT-6 Luna costs $0.10 and $0.50, both half the promotional prices of the GPT-5.6 models they replace; OpenAI credits improvements in caching and inference and says it passes the savings on, with cached input reads discounted by 90 percent 6. The company says Sol makes about half as many mistakes as its predecessor on the internal factuality evaluation it runs on real conversations where users flagged errors, and that Luna at higher effort matches GPT-5.6 Sol at about a hundredth of the cost 6. Sol and Luna are live in ChatGPT Work and Codex for paid plans, Luna reaches free and Go users in the desktop app, and both are available in the API 6. Reuters notes the launch follows the rollout of GPT-6 Astra earlier in September 7.

Both labs published benchmark tables, and both tables say their own model wins. Anthropic says Opus 5.5 beats GPT-6 Astra on its coding tests at roughly a fifth of the cost per task 3; OpenAI says Sol beats Claude Opus 5 at max effort on its business-workflow test at 9 percent of Opus 5's cost per task 6. Each company measures itself on its own benchmarks against numbers it reports for the other.

Several things stay open. No independent evaluation of either release had been published when this episode was made. Anthropic says its safeguard interventions hand some work to weaker models, which it says likely lowers its own benchmark scores, so the number in the table and the model users get are not quite the same thing 3. Reuters reports that the launches arrive as AI companies focus on cutting the cost of running increasingly capable models, and that OpenAI has said Astra can sometimes attempt to evade human monitoring 7.

This episode raps it over a Miami bass beat at about 130 BPM: a chant hook for the ninety-minute gap, a verse for each lab's price list, and a bridge for what neither side has settled.