Price War Erupts: OpenAI and Anthropic Slash Rates as Chinese AI Models Gain Ground

Key Highlights

• OpenAI slashed GPT-5.6 Luna pricing by 80% while reducing Terra costs by 20%
• Anthropic introduced Claude Opus 5 at 50% less than Fable 5, its prior flagship offering
• Anthropic scrapped an upcoming price increase for Claude Sonnet 5
• Major enterprises including DoorDash, Airbnb, and Siemens are experimenting with budget Chinese AI from DeepSeek and Moonshot AI
• Chinese AI solutions have surpassed Claude and ChatGPT in token consumption on OpenRouter

A significant pricing shake-up is underway as OpenAI and Anthropic slash costs across multiple AI offerings, responding directly to enterprise clients exploring more economical Chinese options from providers like DeepSeek and Moonshot AI.

The Economics Behind the Price Drop

Corporate AI spending has entered a new phase of scrutiny. Where budgets were once relatively open-ended, organizations now demand justification for every dollar spent. Chinese developers recognized this opportunity, delivering competitive functionality at dramatically reduced rates for common applications including code generation and customer service automation.

The pricing differential is substantial. Processing one million output tokens through Anthropic’s Fable model runs $50. Moonshot AI’s Kimi K3 handles the same workload for $15. DeepSeek V4 Pro completes it at merely $0.87. For procurement teams managing large-scale deployments, these differences matter.

OpenAI’s response included an aggressive 80% reduction for GPT-5.6 Luna, bringing input token costs from $1 down to $0.20 per million. The Terra tier saw a more modest 20% decrease. While the premium Sol model maintained its existing price point, it gained enhanced processing speed.

Anthropic followed with competitive adjustments. Claude Opus 5 debuted at $5 per million input tokens and $25 per million output tokens—exactly half the pricing of Fable 5, which previously held the top position in their lineup. Additionally, a September price hike scheduled for Sonnet 5 was abandoned, preserving the initial launch pricing.

Enterprise Migration Patterns

Adoption of Chinese AI platforms is no longer theoretical—it’s happening across Fortune 500 companies. DoorDash has integrated Moonshot AI into development workflows. Both Airbnb and Siemens are running trials with solutions from DeepSeek and Alibaba.

OpenRouter platform analytics reveal that Chinese providers DeepSeek and Z.ai have surpassed both Claude and ChatGPT in total token processing volume. DeepSeek specifically claimed the number one position for token-based usage metrics throughout this measurement window.

However, lower per-token pricing doesn’t automatically translate to reduced overall expenditure. A superior model might accomplish identical objectives using fewer tokens while generating fewer mistakes. Industry analysts emphasize that cost-per-completed-task represents a more accurate efficiency metric than simple per-token pricing.

Regardless of whether Chinese alternatives achieve complete market displacement, their availability fundamentally strengthens buyer negotiating position with established American providers.

The competitive pressure facing OpenAI and Anthropic is undeniable. Both organizations carry substantial infrastructure expenses for developing and operating sophisticated models, creating tension as they simultaneously reduce revenue per transaction to maintain market share. The sustainability of these price reductions remains uncertain as usage volumes continue expanding.

From an enterprise perspective, this dynamic delivers expanded options and improved economics. For the AI companies driving innovation, the competitive landscape has fundamentally shifted. Superior capability alone no longer guarantees success—affordability has become equally strategic.