OpenAI and Anthropic Slash Mid-Tier Prices to Combat Chinese AI Rivals
US AI leaders are aggressively discounting mid-range models as enterprises turn to low-cost open-weight alternatives from China.
OpenAI and Anthropic have entered a sharp pricing war for their mid-tier AI models to stem the loss of market share to Chinese competitors. The move signals a strategic shift in the AI industry, moving away from a pure race for raw capability toward a battle over the cost per completed task.
OpenAI reduced the pricing of GPT-5.6 Luna by 80% on July 30, 2026. The model's cost dropped from $1 to $0.20 per million input tokens and from $6 to $1.20 per million output tokens. Simultaneously, Anthropic launched Claude Opus 5 on July 24, 2026, pricing it at $5 per million input and $25 per million output tokens. This represents exactly half the cost of Anthropic's flagship Claude Fable 5, which remains priced at $10 per million input and $50 per million output tokens.
The Rise of Open-Weight Rivals
This pricing aggression is a direct response to the growing influence of Chinese AI firms, specifically Moonshot with its Kimi K3 and DeepSeek. These rivals are gaining ground by offering high-performance, open-weight models. Unlike the closed-API ecosystems of Silicon Valley, open-weight models allow developers to host their own systems, effectively removing API costs and addressing corporate data residency concerns.
The shift is already impacting the enterprise sector. Major companies, including Siemens, Airbnb, and DoorDash, are reportedly testing or adopting these Chinese models to lower their AI operational expenses. As Mantas Lukauskas, AI tech lead at Hostinger, noted, "The US labs have cut the middle and are defending the top."
A Strategic 'Barbell' Approach
Industry analysts describe this as a "barbell" strategy. By aggressively discounting the mid-tier, OpenAI and Anthropic aim to prevent customer churn among developers who prioritize cost. Meanwhile, they maintain high margins on their most advanced frontier models to justify the trillion-dollar valuations associated with their companies.
This suggests that frontier-level intelligence is commoditizing faster than many predicted. If mid-tier models become capable enough to handle the vast majority of enterprise workflows, the demand for ultra-expensive flagship models could shrink. Such a trend would potentially threaten the long-term revenue projections and IPO prospects for the leading US labs.
What to Watch
Market observers are now monitoring whether this price war will lead to a permanent collapse in API margins across the industry. While the pricing for GPT-5.6 Luna and Claude Opus 5 is now confirmed, it remains to be seen if Chinese firms will respond with further performance leaps or deeper discounts. The primary question for the remainder of 2026 is whether the "frontier" premium can be sustained as the gap between mid-tier and flagship intelligence continues to narrow.