OpenAI, Anthropic Slash AI Model Prices Amid Fierce Competition

The artificial intelligence model market is shifting focus from capability to cost, with major players OpenAI and Anthropic announcing significant price reductions for their advanced AI models on the same day. OpenAI launched GPT-6 Sol and GPT-6 Luna with substantially lower prices than their predecessors, just an hour and a half after Anthropic unveiled Claude Opus 5.5. Both companies are signaling increased power at a reduced cost.

OpenAI's Sol is now priced at $2 per million input tokens and $10 per million output tokens, down from $4 and $20 respectively. Luna saw a price drop to $0.10 for input and $0.50 for output, compared to the previous $0.20 and $1.20. The flagship GPT-6 Astra model's price remains unchanged at $10 for input and $50 for output per million tokens. Anthropic's Opus 5.5 is priced at $4 per million input tokens and $20 per million output tokens, a 20% decrease from Opus 5. Cache reads, crucial for programming tasks and long-running agents, are now $0.20 per million tokens. Anthropic claims Opus 5.5 is about 40% cheaper and 30% faster for typical workloads.

The competition extends beyond just token prices, with companies increasingly evaluating the cost to complete a specific task. A cheaper model that fails to achieve the desired outcome can be more expensive than a pricier one that succeeds on the first try. OpenAI highlighted Sol's performance on Zapier's AutomationBench, achieving 33.2% success at $0.27 per task, while claiming Opus 5 performed worse at a significantly higher cost. Anthropic, conversely, presented data showing Opus 5.5 outperforming GPT-6 Astra in several categories.

This price war is not limited to OpenAI and Anthropic. Within 48 hours, xAI released Grok 4.7 at $2 per million input tokens and $6 per output, and Xiaomi introduced its MiMo-V2.6 series under an MIT license, claiming competitive performance for agent tasks.

While lower token prices may seem beneficial, they don't guarantee reduced overall AI expenditures for companies. Increased affordability could lead to more agent activations, more attempts at tasks, and the delegation of work previously done by humans. A company using four times more tokens at half the price might end up with a higher bill. Additionally, OpenAI's Sol and Luna models have higher costs for extremely long prompts exceeding 272,000 tokens, which could impact organizations handling large document analyses or extensive codebases.

For Israeli companies building products on these models, the price drops can improve profit margins but also lower the barrier to entry for new competitors. The advantage is shifting from the ability to afford AI models to the capability of building compelling products around them.