AI slowdown? OpenAI and Anthropic launch cheaper AI models days after urging caution on frontier AI

OpenAI and Anthropic on Tuesday launched lower-cost AI models, expanding their model lineups as businesses increasingly seek affordable AI tools. OpenAI launched GPT-6 Sol and GPT-6 Luna, while Anthropic introduced Claude Opus 5.5. According to the companies, the new models are designed primarily for everyday business tasks rather than the most advanced frontier AI applications.

The launches came within hours of each other and follow recent calls by both companies for a more measured approach, or “pacing,” to frontier AI development. The timing highlights a growing distinction between frontier AI research and commercial AI offerings, with competition increasingly centred on cost, efficiency and enterprise adoption rather than capability alone.

GPT-6 Sol, GPT-6 Luna and Claude Opus 5.5 target lower-cost AI use cases

In a blog post, OpenAI said that GPT-6 Sol is designed for more complex tasks, including coding, while GPT-6 Luna is well-suited for high-volume workloads such as information extraction and document summarization.

The company said API pricing for the two models is 50% cheaper than the promotional rates of the GPT-5.6 models. According to OpenAI, the reductions stem from optimizations in caching and inference.

According to Anthropic, Claude Opus 5.5 is the first model in its new 5.5 family. The company said the model delivers performance comparable to Claude Fable 5.1 on most tasks while costing 40% less to operate than Opus 5.1.

Anthropic priced the model at $4 per million input tokens and $20 per million output tokens and said Sonnet 5.5 and Haiku 5.5 would follow in the coming weeks.

The launches also coincide with the broader shift in enterprise AI spending. According to Fortune, companies are increasingly scrutinising the ROI of AI deployments and opting for lower-cost models to complete tasks. The publication cited Ramp economist Ara Kharazian as saying competition from OpenAI and Anthropic resembles a price war that might squeeze model profitability.

Safety remains critical to Anthropic's model releases

In its blog post, Anthropic said Claude Opus 5.5 incorporates the same protection measures used in its more capable Fable 5.1 model. The company also introduced additional safeguards to prevent model-distillation attacks, in which users attempt to replicate a model's capabilities by repeatedly querying it.

According to Anthropic, external organizations, including METR and Frontier Design, evaluated Opus 5.5 before release. The company also reported its highest score to date on an internal behavioural audit. These findings are company-reported and do not independently establish that all model risks have been eliminated.

'AI slowdown' debate centres on frontier models

The launches also provide context for recent statements about the slowdown in AI development. Dario Amodei, Anthropic's CEO, has argued that the development of highly capable AI systems should be paced gradually to allow safety measures to keep up with increasing capabilities.

OpenAI CEO Sam Altman has expressed similar concerns about frontier AI development. In an August blog post, OpenAI said it had paused some reinforcement-learning training runs, strengthened monitoring and isolation measures and delayed a planned frontier-model training run pending further evaluation.

Taken together, the launches suggest that concerns about AI safety have not slowed competition across the broader market. Instead, the focus appears to be shifting towards tighter controls around frontier AI systems, while companies continue to compete aggressively on pricing, efficiency and commercial adoption.