Anthropic, OpenAI launch new flagship AI models after calling to slow down AI development

Anthropic and OpenAI have released new flagship AI models, Claude Opus 5.5 and GPT-6 Sol respectively. The new launches come at a time when both companies have pushed for a slowdown in AI development.The two companies have released new AI models. (Photos: Reuters)The AI industry is facing one of its biggest challenges right now as companies discuss potentially slowing down models over fears that the technology may be a threat to humanity. Anthropic and OpenAI – two companies leading this push – have now released their latest flagship AI models called Claude Opus 5.5 and GPT-6 Sol respectively. OpenAI also released GPT-6 Luna, a smaller model that is cheaper to run compared to Sol.The two launches centred on a similar pitch – stronger models at lower prices. Anthropic says Opus 5.5 performs at the level of its frontier AI model Claude Fable 5.1 on most work while costing 40 per cent less to run than Opus 5, with input and output token prices cut to $4 (roughly Rs 380) and $20 (roughly Rs 1,900) per million from $5 (roughly Rs 475) and $25 (roughly Rs 2,390).On the other hand, as per OpenAI, GPT-6 Sol and GPT-6 Luna are 50 per cent cheaper than the GPT-5.6 Sol and Luna series, which it attributed to improvements in caching and inference. OpenAI added that the new models make fewer factual and coding mistakes.What’s new with Claude Opus 5.5?Anthropic described Opus 5.5 as the first release in its Claude 5.5 family and said it is its new leading model for tasks such as coding and computer use. The company stated that Opus 5.5 even outpaced the larger Fable model on many benchmarks and completed some informal tasks that Fable failed to finish. Anthropic launches Claude Opus 5.5. Anthropic’s published benchmarks show that Opus 5.5 scored 66.4 per cent on the coding benchmark Terminal-Bench 4.0. For context, Claude Fable 5.1 scored 55.8 per cent on the same benchmark, while OpenAI’s GPT-6 Astra got a score of 57.9 per cent.On the knowledge work test GDPval-AA v2.1, Opus 5.5 scored 1,846 points, compared to 1,735 by Fable 5.1 and 1,542 by GPT-6 Astra. Anthropic says the new AI model writes more clearly while using less jargon. It also uses less computing power and generates output more than 30 per cent faster than Opus 5. In its blog post, Anthropic addressed that Opus 5.5 was its first model since CEO Dario Amodei pushed for “pacing the frontier” when it comes to AI development. Opus 5.5 was tested before release by outside evaluators including METR and Frontier Design. The company says it is the strongest-performing model it has tested so far on its automated behavioural audit for alignment.The release came just two months after Opus 5 launched on 24 July. The company also plans to release lower-cost Sonnet 5.5 and Haiku 5.5 in the coming weeks with similar gains in performance, efficiency and safety.OpenAI’s GPT-6 Sol and Luna join GPT-6 AstraOpenAI’s new models – GPT-6 Sol and GPT-6 Luna join GPT-6 Astra, which the company calls its most powerful model yet. As per OpenAI, Sol is aimed at more complex tasks such as coding, while Luna is meant for high-volume clerical work such as summarising documents or extracting information. OpenAI's GPT-6 family now has three AI models. GPT-6 Sol will cost $2 (roughly Rs 190) per million input tokens and $10 (roughly Rs 955) per million output tokens, down from $4 (roughly Rs 380) and $20 (roughly Rs 1,900) for GPT-5.6 Sol. OpenAI says the two models perform better across business workflows, factuality, coding and computer use. In its internal factuality evaluation based on de-identified conversations in which users flagged mistakes, the company stated that GPT-6 Sol makes about half as many mistakes as its predecessor and reaches Astra-level reliability at much lower cost.In a blog post, OpenAI shared that GPT-6 Sol could score 56.4 per cent on Agents’ Last Exam at max effort, higher than Claude Fable 5. On OSWorld 2.0 offline, GPT-6 Sol at xhigh effort reached 60.5 per cent against 60.3 per cent for Claude Opus 5 at medium effort, at roughly 80 per cent lower cost per task.OpenAI also confirmed that GPT-6 Sol and GPT-6 Luna were its most aligned AI models to date.The AI slowdown debate followed a summer in which concerns about AI safety sharpened. Ever since about 700 rogue OpenAI agents tried to hack Hugging Face, many are concerned that AI may be a threat to humanity. Some, like Anthropic’s Evan Hubinger, have predicted that AI may kill all humans within the next decade. Such fears have led to Anthropic’s Dario Amodei and OpenAI’s Sam Altman pushing for regulation around AI development.Though so far, Silicon Valley seems divided. Nvidia CEO Jensen Huang and Meta CEO Mark Zuckerberg have pushed back against such an idea. US President Donald Trump has also rejected such proposals for now.- EndsPublished By: Armaan AgarwalPublished On: Sep 23, 2026 08:54 IST