New AI Models That Are Cheaper and Make Fewer Mistakes
Anthropic and OpenAI both launched new AI models on September 22, 2026 that they say work better and cost less.
Anthropic introduced Opus 5.5. OpenAI introduced GPT-6 Sol and GPT-6 Luna. Engadget OpenAI's news index lists 'Introducing GPT-6 Sol and Luna' as a Product post dated September 22, 2026. Anthropic's news index lists the Claude Opus 5.5 announcement on the same date.
Opus 5.5 costs $4 for input tokens and $20 for output tokens, compared with $5 for input and $25 for output for Opus 5. Engadget Tokens are small chunks of words the model reads and writes. That is a 20% drop in list API price, the published price for developers. In separate materials Anthropic says Claude Opus 5.5 costs 40% less to run than Opus 5 and performs at the level of Claude Fable 5.1 on most work. Opus 5.5 is available to developers through Claude, Amazon Web Services, Google Cloud and Microsoft Azure. Engadget
Anthropic says Opus 5.5 tried to get around its safety limits about 85 percent less often than past models. Engadget On skills, Anthropic's test results claim Opus 5.5 did better at coding jobs than GPT-6 Astra on Terminal-Bench 4.0 and FrontierCode v1.1 (Main). Those tests measure multi-step coding work in a computer terminal, not single answers. Those numbers come from the vendor. They have not been independently reproduced in the verified facts.
Price, performance and guardrails
OpenAI says GPT-6 Sol and GPT-6 Luna were trained using similar methods to GPT-6 Astra. Engadget Pricing splits the line into a mid-tier workhorse and a low-cost high-volume option. GPT-6 Sol is priced at $2 for input tokens and $10 for output tokens. GPT-6 Luna is priced at $0.10 for input tokens and $0.50 for output tokens.
OpenAI says GPT-6 Sol makes about half as many factual mistakes as its predecessor. The verified facts do not specify the test behind that claim. For uses where AI looks up documents before answering, uses tools, or sums up long texts, accuracy matters more than top test scores. It affects checking work, repeat runs, and human review cost. A halving, if it holds in real use, changes running costs even before price cuts are counted.
The price levels point to different jobs. Luna pricing sits two orders of magnitude below Opus-class pricing. Listed uses include sorting, routing, supervision of smaller models, bulk summaries, and small agent subtasks where a larger model makes the plan and smaller models do the steps. Think of a foreman with a work plan and a crew handling routine tasks. Sol sits between Luna and Opus 5.5. It is cheap enough for high-throughput coding help and company search, while keeping the GPT-6 training lineage.
Where the models ship
Access paths differ. Anthropic offers Opus 5.5 direct and through large cloud providers at once: Claude plus AWS, Google Cloud and Azure. That lets workloads stay inside existing cloud deals, VPC controls, and buying systems. VPC controls are private network and permission settings inside a company cloud account. It also keeps computing near customer data gravity, where company data already lives.
OpenAI made GPT-6 Sol and GPT-6 Luna available in ChatGPT Work and Codex for Plus, Pro, Business and Enterprise customers, with Free users and Go subscribers getting only GPT-6 Luna in the desktop app. Engadget Paid tiers get both models for work and coding flows. The free and entry tier gets Luna on desktop, which limits costly computing while spreading use.
The launch history behind both includes many steps. Anthropic announced the Claude 3 family on March 4, 2024, Claude Opus 4 on May 22, 2025, Claude Opus 4.5 on November 24, 2025, Claude Opus 4.8 on May 28, 2026, and Claude Opus 5 on July 24, 2026. Anthropic states Claude Opus 4 leads on SWE-bench at 72.5% and Terminal-bench at 43.2%, that Claude Opus 4.8 can work at 2.5x the speed, and that Claude Opus 5 is the strongest Opus model it has tested on its trading benchmark. It announced Claude Fable 5.1 and Claude Mythos 5.1 on September 1, 2026. OpenAI's prior price-performance steps include GPT-4o mini on July 18, 2024, described as its most cost-efficient small model and more than 60% cheaper than GPT-3.5 Turbo, and GPT-5.6 Luna on July 30, 2026, described as its fastest and most affordable model at 80% less cost. It also previewed GPT-5.6 Sol as a next-generation model and said GPT-5.6 Terra has competitive performance to GPT-5.5 while being 2x cheaper.
Immediate context helps explain timing. CNBC reported on September 18, 2026 that Anthropic and OpenAI were hunting for smaller AI data center deals in a race to deploy AI capacity. Reuters reported on September 18, 2026 that Anthropic was considering rolling out a new AI model to counter OpenAI's momentum since its launch of GPT-6 Astra. OpenAI launched GPT-6 Astra, according to that same reporting.
The broader context here is a change from making models bigger in training to making them cheaper to run. Both labs now announce price cuts with better performance claims. That was not always the pattern. Earlier top launches led with test wins first, with cheaper versions later in mini or distilled variants. Sol, Luna and Opus 5.5 combine those steps. Low cost is part of the launch message.
In my view, teams should read this less as a race and more as a signal about daily operations. When vendors cut list prices 20% to 80% while claiming fewer safety workarounds and fewer factual errors, they aim for AI helpers used at large scale. The hard parts shift to managing workflows, test setups, reuse rates, and fixing errors when agents act on wrong outputs.
Worth flagging is that company tests on Terminal-Bench 4.0, FrontierCode v1.1, SWE-bench and trading tasks help with first sorting, but they do not replace testing with your own code, permissions and data.
What this makes possible, if the claims hold in practice, is simple. Cheaper, steadier models make it affordable to let agents run longer, do more tasks at once, and add extra checks once too costly. That favors a strong planner, cheap helpers, and clear checkers. Over time, that is how new tools turn into reliable basics. Misuse risks remain, as Anthropic's September 2026 threat intelligence report notes in covering seven areas of harm from December 2025 to August 2026, but the path is toward more useful automation at lower cost.
