The battle for enterprise AI dominance has triggered a massive AI price war, with US giants OpenAI and Anthropic slashing the costs of their mid-tier models to fend off aggressive Chinese competitors. As models like DeepSeek's V4 Flash and Moonshot's Kimi K3 gain ground, the trillion-dollar ambitions of Silicon Valley are facing a severe reality check. Customers are increasingly scrutinizing the cost of input and output tokens, forcing industry leaders to adjust their pricing strategies to maintain market share.
OpenAI recently reduced the cost of its GPT-5.6 Luna model by a staggering margin, dropping the price from $1 to $0.20 per million input tokens, and from $6 to $1.20 per million output tokens. Anthropic quickly followed suit, launching Opus 5 at $5 per million input tokens and $25 per million output tokens - exactly half the cost of its flagship Fable 5 model. Furthermore, Anthropic canceled a planned September price hike for its Sonnet 5 model.
However, an insider close to Anthropic pushed back on the narrative that these cuts were purely reactive. The source stated that pricing Opus 5 below Fable 5 is simply how the startup's "family of models is built, so there's no connection to competitors."
Benchmarking the Global Competition
Headline token prices do not always tell the full story when evaluating an AI price war. The ultimate cost of a task depends heavily on a model's efficiency and its "effort" settings, which dictate the computing power allocated to a prompt. More capable models can sometimes complete complex reasoning tasks using fewer tokens or fewer attempts, altering the true cost-per-task.
According to data from Artificial Analysis, Anthropic's Opus 5 running at "medium" effort delivers similar performance and cost-per-task to Moonshot's Kimi K3 operating at "max" effort. Meanwhile, OpenAI's GPT-5.6 Luna at "max" effort matches the performance of DeepSeek's V4 Flash at "max," but the OpenAI model costs nearly twice as much per task.
The Squeeze on the Middle Market
Mantas Lukauskas, AI tech lead at Hostinger, noted that while mid-tier prices are plummeting, the costs for the absolute best models remain "flat to rising." He views these recent pricing changes as the first real test of whether groups like Anthropic and OpenAI can protect the profitability of their most advanced offerings.
The US labs have cut the middle and are defending the top.
- Mantas Lukauskas, AI Tech Lead, Hostinger
This dynamic reveals a clear bifurcation in the AI market. By commoditizing their mid-tier offerings, OpenAI and Anthropic are attempting to starve out cheaper international rivals while locking enterprise clients into their ecosystems. Once developers build their infrastructure around GPT-5.6 Luna or Opus 5, the friction to switch to DeepSeek becomes a significant barrier.
This ecosystem lock-in allows US labs to maintain premium pricing for their cutting-edge flagship models, where true enterprise value is generated. If Chinese labs want to break this defense, they will need to prove their models can not only compete on price in the middle tier, but also match the reasoning capabilities of the industry's most expensive top-tier systems.
