Key Points
- Anthropic released Claude Opus 5.5 on Sept. 22, 2026, the first model in its Claude 5.5 family, with company-stated pricing of $4 input / $20 output per million tokens.
- Anthropic says Opus 5.5 costs about 40% less on typical workloads than Opus 5, with cache reads at $0.20 per million tokens and more than 30% faster output (company claims).
- OpenAI followed within about 90 minutes with GPT-6 Sol and GPT-6 Luna, expanding the GPT-6 line after Astra (TechCrunch).
- OpenAI says API prices are about 50% below GPT-5.6 Sol/Luna promotional pricing; Sol targets complex work such as coding, while Luna targets high-volume clerical tasks and reaches Free/Go users on desktop (company / TechCrunch).
- Both launches land after the companies leaders publicly urged pacing frontier AI development; treat performance and price advantages as vendor claims.
Sept. 22, 2026 delivered a same-day duel at the top of the U.S. AI stack. Anthropic launched Claude Opus 5.5. About 90 minutes later, according to TechCrunch, OpenAI expanded its GPT-6 family with GPT-6 Sol and GPT-6 Luna. Searchers looking up Claude Opus 5.5 OpenAI GPT-6 want the product facts, the price claims, and the competitive timing without treating every lab scoreboard as settled science.
Anthropic ships Claude Opus 5.5
Anthropics product post introduces Claude Opus 5.5 as the first model in its Claude 5.5 family. The company says the model performs at the level of Claude Fable 5.1 on most work while costing less to run than Opus 5. External pre-release evaluators cited by Anthropic include Frontier Design and METR.
On pricing, Anthropic lists $4 per million input tokens and $20 per million output tokens, which it describes as 20% below Opus 5’s $5 / $25 rates. Cache reads are $0.20 per million tokens, 60% below Opus 5’s $0.50 figure. Anthropic’s own tests claim about 40% lower cost than Opus 5 on typical workloads, plus more than 30% faster output. Those figures are company claims, not independent LiveNewsWorld benchmarks.
Availability spans Anthropic’s platforms plus Amazon Web Services, Google Cloud, and Microsoft Azure. Developers can call the model as claude-opus-5-5. Anthropic also says biology and cybersecurity safeguards are similar in class to those on Fable 5.1, with Life Sciences and Cyber Verification programs for vetted access. This package stays product-focused and does not expand into unverified misuse scenarios.
OpenAI answers with GPT-6 Sol and Luna
TechCrunch reports that OpenAI launched updated GPT-6 Sol and GPT-6 Luna models the same day, expanding the GPT-6 generation after GPT-6 Astra earlier in the month. Sol is positioned for complex tasks such as coding. Luna is positioned for high-volume clerical work with a clear goal, such as summarizing documents, extracting information, or answering quick questions.
OpenAI attributes a substantial API price drop to caching and inference gains and says the GPT-6 Sol/Luna series will be available at about half the cost of the GPT-5.6 Sol/Luna promotional pricing. That ~50% cheaper framing is an OpenAI promotional claim and should be labeled as such. OpenAI also claims that on an internal factuality evaluation based on de-identified conversations where users flagged mistakes, GPT-6 Sol makes about half as many mistakes as its predecessor and approaches Astra-level reliability at lower cost. Again: vendor claim.
Rollout details in TechCrunch: Sol and Luna are available in ChatGPT Work and Codex for most paid accounts and in the ChatGPT API; Luna is also available in the desktop app for Free and Go users; ChatGPT web and app rollout is gradual.
Why the ~90-minute gap matters
TechCrunch explicitly notes that Anthropic’s Opus 5.5 release landed about 90 minutes before OpenAI’s Sol/Luna announcement. That is the competitive beat of the day: two labs shipping cheaper, faster companions around the same news cycle. Anthropic’s own post frames Opus 5.5 as its first release since CEO Dario Amodei called for pacing the frontier. Secondary coverage notes that both labs have recently urged caution after summer security concerns, then still shipped product. LiveNewsWorld reports that tension as attributed industry context, without moralizing who is right.
How to read the claims
Readers should separate three layers:
- Confirmed shipping facts: Opus 5.5 is live with stated list prices and cloud availability; Sol and Luna are live with stated role splits and rollout channels.
- Vendor performance and savings claims: ~40% lower typical cost, >30% faster output, ~50% API discount vs GPT-5.6 promo pricing, and “half as many mistakes” are company statements.
- Unconfirmed extrapolation: Head-to-head “winner rankings beyond what each lab publishes, exact consumer ChatGPT timing beyond gradual, and detailed safety-incident reconstructions belong in separate reporting with primary documents.
Anthropic publishes comparative tables that include rival figures as reported by OpenAI and other evaluators. Those tables are useful context, but competitor numbers on a lab’s own page remain attributed data. Independent verification of every percentage is outside this package.
Enterprise and consumer takeaways
For API and agent buyers, the same-day message is price pressure at the near-frontier tier. Anthropic is pitching Opus 5.5 as near Fable-class work at Opus-discounted economics. OpenAI is pitching Sol as a cheaper path toward Astra-like reliability for hard tasks, and Luna as the high-volume clerical tier that reaches Free/Go desktop users.
For broader AI market context already on LiveNewsWorld, see AI and tech remain a major market story, Nvidia earnings live, and the site’s AI category.
Bottom line
Claude Opus 5.5 OpenAI GPT-6 coverage on Sept. 22 is a dual product launch story: Anthropic cut Opus pricing and raised speed claims; OpenAI answered ~90 minutes later with cheaper Sol and Luna tiers. Publish the shipping facts. Label the savings and quality percentages as vendor claims. Keep the safety-slowdown irony attributed, not sermonized.
