The AI industry is currently dealing with one of its biggest challenges, with companies discussing whether they should slow the development of AI models amid concerns that the technology could pose a danger to humanity. Anthropic and OpenAI, two companies at the centre of this debate, have now launched their newest flagship models, Claude Opus 5.5 and GPT-6 Sol. OpenAI has also introduced GPT-6 Luna, a smaller and more cost-efficient model compared with Sol.
Both launches focused on a similar proposition: more capable AI models that cost less to operate. Anthropic said Opus 5.5 delivers performance comparable to its frontier model, Claude Fable 5.1, for most tasks while being 40 per cent cheaper to run than Opus 5. Its input and output token prices have been reduced to $4 (around Rs 380) and $20 (around Rs 1,900) per million, from $5 (around Rs 475) and $25 (around Rs 2,390), respectively.
OpenAI, meanwhile, said GPT-6 Sol and GPT-6 Luna are 50 per cent cheaper than the GPT-5.6 Sol and Luna models, crediting improvements in caching and inference. The company also said the new models produce fewer factual and coding errors.
What’s new with Claude Opus 5.5?
Anthropic called Opus 5.5 the first model in its Claude 5.5 series and described it as its new leading AI system for tasks including coding and computer use. According to the company, Opus 5.5 surpassed the larger Fable model on several benchmarks and was able to complete some informal tasks that Fable could not.
[Claude Opus 5.5]
Anthropic launches Claude Opus 5.5.
According to Anthropic’s published benchmark results, Opus 5.5 achieved 66.4 per cent on the Terminal-Bench 4.0 coding test. By comparison, Claude Fable 5.1 scored 55.8 per cent, while OpenAI’s GPT-6 Astra recorded 57.9 per cent.
In the GDPval-AA v2.1 knowledge-work evaluation, Opus 5.5 scored 1,846 points, ahead of Fable 5.1’s 1,735 and GPT-6 Astra’s 1,542. Anthropic also said the model produces clearer writing with less jargon. It requires less computing power and generates responses more than 30 per cent faster than Opus 5.
In a blog post, Anthropic noted that Opus 5.5 is the first model released since CEO Dario Amodei called for “pacing the frontier” in AI development. Before its launch, the model was evaluated by external groups including METR and Frontier Design. Anthropic said Opus 5.5 is the strongest-performing model it has so far assessed through its automated behavioural alignment audit.
The launch comes only two months after Opus 5 was released on July 24. Anthropic is also expected to introduce the lower-cost Sonnet 5.5 and Haiku 5.5 in the coming weeks, with the company promising improvements in performance, efficiency and safety.
OpenAI’s GPT-6 Sol and Luna join GPT-6 Astra
OpenAI’s latest models, GPT-6 Sol and GPT-6 Luna, now join GPT-6 Astra, which the company describes as its most powerful model so far. OpenAI said Sol is designed for more demanding tasks such as coding, while Luna is intended for high-volume administrative work, including document summarisation and information extraction.
[Anthropic]
OpenAI's GPT-6 family now has three AI models.
GPT-6 Sol is priced at $2 (around Rs 190) per million input tokens and $10 (around Rs 955) per million output tokens. This is down from $4 (around Rs 380) and $20 (around Rs 1,900), respectively, for GPT-5.6 Sol.
OpenAI said both new models deliver stronger results in areas including business workflows, factual accuracy, coding and computer use. In an internal factuality assessment based on de-identified conversations where users had reported mistakes, the company said GPT-6 Sol made roughly half as many errors as its predecessor and achieved reliability comparable to Astra at a substantially lower cost.
In a blog post, OpenAI said GPT-6 Sol achieved 56.4 per cent on Agents’ Last Exam at maximum effort, exceeding Claude Fable 5’s score. On the offline OSWorld 2.0 benchmark, GPT-6 Sol at xhigh effort scored 60.5 per cent, compared with 60.3 per cent for Claude Opus 5 at medium effort, while costing around 80 per cent less per task.
OpenAI also said GPT-6 Sol and GPT-6 Luna are the company’s most aligned AI models to date.
The debate over slowing AI development intensified during a summer when concerns over AI safety grew. After around 700 rogue OpenAI agents reportedly attempted to hack Hugging Face, concerns increased over the possibility of AI posing a threat to humanity. Anthropic safety researcher Evan Hubinger has also warned that AI could potentially kill all humans within the next decade. These concerns have prompted Anthropic CEO Dario Amodei and OpenAI CEO Sam Altman to advocate for regulation of AI development.
However, Silicon Valley remains divided on the issue. Nvidia CEO Jensen Huang and Meta CEO Mark Zuckerberg have opposed the idea of slowing AI development. US President Donald Trump has also rejected such proposals for the time being.
*-
