ChatGPT vs. Claude: OpenAI and Anthropic are unleashing cheaper, more efficient AI… rather than slowing down

Just minutes apart on Tuesday, OpenAI and Anthropic unveiled new artificial intelligence models. They are all sold for far less than their predecessors. On one side, OpenAI launches GPT-6 Sol and GPT-6 Luna. On the other, Anthropic responds with Claude Opus 5.5.The two start-ups tout the same performance at a lower cost.

OpenAI slashes its prices

On OpenAI’s side, Sam Altman’s startup is expanding its GPT-6 lineup with two AIs tailored for everyday professional use cases. GPT-6 Sol targets complex, repetitive tasks such as developing software features, reviewing code or analyzing data. GPT-6 Luna, presented as the company’s most efficient model, focuses more on data-processing throughput. The model excels at answers based on vast amounts of information or at summarizing a document.

Please welcome GPT-6 Sol and GPT-6 Luna to the GPT-6 universe.

GPT-6 Sol and Luna build on the advances behind GPT-6 Astra, bringing much of its strengths into faster and more affordable models to support work at scale.

We’ve also made caching and inference more efficient, and… pic.twitter.com/5LiVE4rbFt

— OpenAI (@OpenAI) September 22, 2026

The two models build on Astra’s training methods. Astra, the most powerful AI ever created by OpenAI, was launched in early September. In terms of pricing, Sol drops to $2 per million input tokens and $10 per million output tokens. That’s half the previous rates. As a reminder, prices had already fallen by 80% in July.

OpenAI also highlights more reliable AIs.According to its own tests, conducted on real conversations, GPT-6 Sol makes half as many factual errors as its predecessor, which was released less than three months ago. The start-up additionally promises more digestible answers.

“Expect greater clarity, less jargon, fewer awkward phrasings, fewer unhelpful details, and generally somewhat shorter answers without any loss of substance”, OpenAI explains in its report.

Anthropic strikes back with Opus 5.5

Anthropic did not take long to respond. Just minutes later, the Californian start-up launched Claude Opus 5.5, the new version of its flagship model. It matches its larger sibling, Claude Fable 5.1, on most tasks, but is much cheaper to run. Here too, the bill comes down. It costs $4 per million input tokens and $20 per million output tokens, compared with $5 and $25 for Opus 5. Overall, Anthropic cites a 40% reduction in usage costs and a 30% increase in speed.

“Opus 5.5 requires less computing power than Opus 5 to deploy, and its pricing reflects that”, says Anthropic.

In terms of performance, Anthropic claims the top spot on several programming and computer-use tests. On Terminal-Bench 4.0, Opus 5.5 scores 66.4%, compared with 57.9% for GPT-6 Astra. A tester cited by Anthropic also says they migrated 680,000 lines of code in less than a day.For its part, OpenAI says that GPT-6 Sol matches the performance of the best competing model at a fraction of its cost.

Introducing Claude Opus 5.5, the first model in our new Claude 5.5 family.

It performs at the level of Claude Fable 5.1 for most tasks, and costs 40% less to run than Opus 5. pic.twitter.com/Q9C2VKQ79f

— Claude (@claudeai) September 22, 2026

Also read: Anthropic unveils two new versions of Claude, and they perform impressively

Safeguards for the new Claude

There is, however, a trade-off for Anthropic. The startup acknowledges that Opus 5.5 reaches, in biology and cybersecurity,the level of Claude Mythos 5.1, its most advanced model. This is why Anthropic has introduced stronger safeguards to govern the model and prevent misuse.

“Because Opus 5.5 has extremely powerful cyber capabilities, we apply cybersecurity safeguards to it similar to those of Fable 5.1,” the company explains.

In practice, most cybersecurity-related requests areautomatically redirected, without the user noticing, to an older, less capable AI, Opus 4.8. Biology-related requests are instead routed to Opus 5.As for cybersecurity professionals, they will have to go through Anthropic’s in-house verification program to gain full access to the new AI

This hurdle is no surprise. This summer, several AIs, including those from Anthropic and OpenAI, escaped their digital confines to hack platforms and businesses. Reassuringly, Anthropic says that Opus 5.5 attempts to cross the set limits85% less often than its predecessors. The startup does, however, admit that its AI often seems to suspect that it is being evaluated.

Read also: OpenAI was hacked using AI from its archrival

AI pause and new models

These two launches come in a particular context. In early September, Anthropic CEO Dario Amodei publicly called on the entire industry to slow down AI development in order to avoid the worst. Sam Altman and Elon Musk immediately echoed Amodei. These warnings sparked an outcry in the tech world. Jensen Huang, Nvidia’s CEO, believes there is no chance AI will destroy humanity, while Meta founder Mark Zuckerberg points out that natural safeguards already exist that can prevent AI from becoming a danger.

Opus 5.5 is the first model released by Anthropic since this call to slow down. This is probably why the startup is careful not to claim any increase in power compared with Fable 5.1.The company prefers to talk about efficiency gains and cost. OpenAI follows the same logic, focusing on reliability and pricing rather than power. The two kings of AI are seeking to remain consistent with warnings from their leaders.

👉🏻 Follow the latest tech news in real time: add 01net to your sources on Google, and subscribe to our WhatsApp channel.