The frontier AI market has transformed into a price war as Anthropic and OpenAI engage in rapid-fire model launches and price cuts. According to Investing.com India, Claude Opus 5.5 arrived around lunchtime in New York carrying what would have been enough ammunition to dominate the conversation for days, but by early afternoon, OpenAI released GPT-6 Sol and GPT-6 Luna with even lower prices. The competition is now shifting from benchmark trophies to cost per completed task, where token efficiency, caching, and agent economics matter as much as raw capability. OpenAI CEO Sam Altman and Tesla and SpaceX CEO Elon Musk joined Anthropic's Dario Amodei's call for an industry-wide slowdown on September 12, but ten days later, Anthropic released Opus 5.5 and OpenAI answered almost immediately with aggressive pricing cuts.
As reported by Business Standard, Anthropic launched Claude Opus 5.5, the first model in its new Claude 5.5 family. The company said the model delivers performance comparable to its more powerful Fable 5.1 model on most work, while costing about 40% less to run than its predecessor, Opus 5. According to The Straits Times, Opus 5.5 matches Anthropic's cutting-edge model, Fable, on most tasks at a 20% lower price. Opus 5.5 is priced at $4 per million input tokens and $20 per million output tokens. The company plans to follow the release with Sonnet 5.5 and Haiku 5.5 in the coming weeks. Anthropic's higher-end model, Fable 5.1, costs $10 per million input tokens and $50 per million output tokens. According to the latest reports, Anthropic claims Opus 5.5 is better at handling complex work, ' and the company's head of product management, research and labs Dianne Penn told CNBC that they are continuing to innovate on making thinking and answering more efficient, using less tokens depending on effort settings.
According to Business Standard, OpenAI released GPT-6 Sol and GPT-6 Luna, lower-cost models built on the technology behind its GPT-6 Astra model, which was launched earlier this month. As reported by The Straits Times, the new models are cheaper, faster versions of GPT-6 Astra, OpenAI's most powerful model, launched earlier in September. The company said Sol and Luna are designed to bring the capabilities of Astra to everyday work while offering faster performance and lower costs. Their API prices are 50% lower than the promotional pricing for the previous GPT-5.6 models, with OpenAI citing improvements in caching and inference efficiency. The new models are available in ChatGPT Work and Codex for Plus, Pro, Business and Enterprise customers, with free users and Go subscribers having access to only GPT-6 Luna in the company's desktop app. According to Investing.com India, Sol now comes in at $2 per million input tokens and $10 per million output, while Luna drops all the way to ten cents in and fifty cents out. OpenAI claims these models handle tasks substantially better than Anthropic's top models.
The AI market is experiencing a 73% collapse in Opus-tier list pricing in a little over a year, with Claude Opus 4.1 charging $15/$75 in August 2025, Opus 4.5 resetting to $5/$25, and GPT-5.6 Sol arriving at $5/$30 before OpenAI knocked it down to $4/$20. As reported by Investing.com India, open-weight models accounted for 56% of token traffic on Vercel's AI Gateway in August, up from just 7% in December, yet represented only 14% of estimated spend. On OpenRouter, open-weight models, mostly Chinese, made up around 60% of US token usage in August, while the arithmetic suggests the average closed-model token was costing nearly eight times as much. Despite these price pressures, Anthropic still captured 64% of spending through Vercel's gateway even as customers traded down the stack, with the company's annualized revenue run rate reaching $65 billion by the end of July, up from $9 billion at the end of 2025.
According to Business Standard, Anthropic CEO Dario Amodei called for the industry to "pace the frontier" of AI development, arguing that AI capabilities should be developed so that safety measures can keep up. OpenAI CEO Sam Altman has also backed this call. The debate about model safety reached a fever pitch in recent weeks, with former Anthropic researcher Jacob Coxon setting off the controversy when he posted to X on September 8 that he had quit his job, warning that both labs were "gambling with our lives." Anthropic claims Opus 5.5 produces writing that's easier to understand and "attempted to circumvent boundaries around 85 percent less often" than past models, suggesting it'll disobey directions less often. The company's benchmarks show Opus 5.5 scored better at agentic coding than GPT-6 Astra on both Terminal-Bench 4.0 and FrontierCode v1.1 (Main), which could make it more appealing for developers. OpenAI says the GPT-6 models are more aligned than before based on the company's homegrown tests, with Sol making about half as many mistakes as its predecessor.
