Key facts
- OpenAI reduced prices for GPT-5.6 Luna to $0.20 per million input tokens and $1.20 per million output tokens.
- Anthropic launched Opus 5 at $5 per million input tokens and $25 per million output tokens.
- Anthropic maintained current pricing for its Sonnet 5 model, canceling a planned increase.
- Benchmarking indicates that OpenAI's GPT-5.6 Luna at max effort costs nearly twice as much per task as DeepSeek's V4 Flash.
- Anthropic's Opus 5 at medium effort performed similarly in cost and performance to Moonshot's Kimi K3 at max effort.
OpenAI and Anthropic have implemented price cuts on some of their artificial intelligence models in an effort to remain competitive with Chinese AI rivals. OpenAI reduced the cost of its GPT-5.6 Luna model, while Anthropic launched its Opus 5 model at a lower price point and decided against increasing prices for its Sonnet 5 model.
These pricing adjustments primarily affect mid-tier products, aiming to make them more appealing compared to offerings from Chinese competitors. However, direct comparisons of AI model costs are complicated by factors such as the number of tokens used to complete a task and the varying "effort" settings that influence computing power and overall expense.
Benchmarking by Artificial Analysis revealed that while OpenAI's GPT-5.6 Luna at maximum effort performed similarly to DeepSeek's V4 Flash, it was nearly twice as expensive per task. Anthropic's Opus 5, when run at a medium effort setting, showed comparable performance and cost to Moonshot's Kimi K3 operating at its maximum effort.
Mantas Lukauskas, AI tech lead at Hostinger, noted that prices for the most advanced models have remained stable or increased. He described the recent price cuts as a crucial test for Anthropic and OpenAI to see if they can maintain the cost of their top-tier offerings, suggesting they are cutting prices in the middle while defending their most advanced products.