# OpenAI Slashes GPT-5.6 Luna Prices 80%, Undercutting DeepSeek as AI Price War Intensifies > Source: > Published: 2026-07-31 12:24:47.721639+00:00 # OpenAI Slashes GPT-5.6 Luna Prices 80%, Undercutting DeepSeek as AI Price War Intensifies - OpenAI cut GPT-5.6 Luna pricing 80% to $0.20/$1.20 per million input/output tokens, down from $1/$6, effective July 30 [[1]](https://the-decoder.com/openai-goes-full-china-pricing-mode-with-an-80-percent-cut-to-its-most-affordable-gpt-5-6-model/) - GPT-5.6 Terra dropped 20% to $2/$12 per million tokens; the flagship Sol model held steady at $5/$30 [[2]](https://finance.yahoo.com/technology/ai/articles/openai-cuts-gpt-5-6-173045044.html) - Luna's new $0.20 input price undercuts DeepSeek V4 Pro's $0.435 per million input tokens, though DeepSeek remains cheaper on outputs [[3]](https://eu.36kr.com/en/p/3919319636290946) - Chinese models have captured 46% of US enterprise token usage on OpenRouter, peaking above US-origin models [[4]](https://www.cnbc.com/2026/07/30/open-ai-price-cut-gpt.html) - DeepSeek launched its V4 model on the same day as OpenAI's price cut, escalating the head-to-head competition [[3]](https://eu.36kr.com/en/p/3919319636290946) OpenAI slashed prices on two of its three GPT-5.6 models on July 30, cutting the cost of its fastest tier, Luna, by 80% and the mid-range Terra by 20%. Luna now costs $0.20 per million input tokens and $1.20 per million output tokens, down from $1 and $6 respectively — a move that undercuts DeepSeek's V4 Pro on input pricing just three weeks after the GPT-5.6 family launched on July 9 [[1]](https://the-decoder.com/openai-goes-full-china-pricing-mode-with-an-80-percent-cut-to-its-most-affordable-gpt-5-6-model/) [2]. The price reductions come amid mounting competitive pressure from Chinese AI providers, which have captured 46% of US enterprise token usage on the OpenRouter platform, according to a CNBC investigation published earlier in July [4]. Microsoft and Google have also been expanding lower-cost model offerings, squeezing OpenAI's pricing power from multiple directions . [[2]](https://finance.yahoo.com/technology/ai/articles/openai-cuts-gpt-5-6-173045044.html)OpenAI attributed the cuts to infrastructure efficiency gains rather than competitive pressure, saying its flagship Sol model autonomously optimized GPU software and production code, reducing end-to-end serving costs by 20% and boosting token-generation efficiency by more than 15% through speculative decoding techniques [[1]](https://the-decoder.com/openai-goes-full-china-pricing-mode-with-an-80-percent-cut-to-its-most-affordable-gpt-5-6-model/) [2]. ## The New Pricing The restructured API pricing creates a sharply tiered lineup. Luna, designed for high-volume and latency-sensitive workloads, now sits at $0.20/$1.20 per million input/output tokens — down 80% from its launch price of $1/$6. Terra, the balanced mid-tier model, dropped 20% to $2/$12 from $2.50/$15. The flagship Sol model remains unchanged at $5/$30 per million tokens [[1]](https://the-decoder.com/openai-goes-full-china-pricing-mode-with-an-80-percent-cut-to-its-most-affordable-gpt-5-6-model/) [2]. OpenAI also introduced a new 'Fast Mode' for the Sol API, delivering 2.5x faster inference speeds at double the standard price, replacing the previous Priority Processing tier [5]. ChatGPT and Codex subscription users benefit indirectly: Terra and Luna now consume fewer credits within existing subscription tiers, though subscription prices themselves remain unchanged [5]. ## Competitive Landscape Luna's $0.20 input price now undercuts DeepSeek V4 Pro, which charges $0.435 per million input tokens with a standing 75% promotional discount. On output tokens, however, DeepSeek remains cheaper at $0.87 versus Luna's $1.20 [3]. The timing was pointed: DeepSeek released its V4 model on the same day as OpenAI's announcement, claiming improved coding and debugging capabilities. Chinese models reached 66.5% of token volume on OpenRouter by late July, up from roughly half in April [3]. Anthropic, OpenAI's closest domestic rival, faces pricing pressure as well. Its Claude Sonnet 5 is currently priced at an introductory $2/$10 per million tokens, rising to $3/$15 after August 31 — both above Luna's new rates. Anthropic's top-tier Fable 5 sits at $10/$50 [2]. ## The Self-Optimization Claim OpenAI framed the cuts as a technology story rather than a competitive reaction. The company said Sol — its most capable model — had been used to rewrite and optimize its own inference infrastructure, autonomously improving GPU utilization and production code [1]. The result, OpenAI claimed, was a 20% reduction in end-to-end serving costs and a 15%-plus improvement in token generation throughput via speculative decoding. The company said tasks that previously cost $1 on Luna now run for approximately 6 cents at nearly nine times the original speed [1]. The claim that a frontier model can meaningfully reduce its own deployment costs represents a notable development in AI infrastructure economics, though independent verification of the efficiency figures has not been published. ## Enterprise Context The aggressive pricing arrives as enterprise customers grow increasingly cost-conscious about AI spending. Companies have been reluctant to authorize large AI expenditures without clear return-on-investment metrics, and some have switched to cheaper alternatives or established internal spending tiers [2]. Both OpenAI and Anthropic have filed confidential IPO prospectuses, creating additional pressure to demonstrate sustainable business models even as they cut prices. OpenAI's dual strategy — preserving Sol's premium positioning while defending mid-tier market share — reflects the tension between revenue growth and competitive positioning [2]. The price war extends beyond API costs. Microsoft has been promoting its own MAI models as cost-effective alternatives for enterprise customers, while Google has expanded its lower-cost Gemini offerings, creating pressure on multiple fronts for frontier AI labs whose balance sheets depend on infrastructure-heavy capital expenditure [[1]](https://the-decoder.com/openai-goes-full-china-pricing-mode-with-an-80-percent-cut-to-its-most-affordable-gpt-5-6-model/) [4]. ## What's Next Luna's new pricing targets high-volume, repetitive tasks and agent-based workflows that were previously dominated by cheaper Chinese models. OpenAI is betting that matching or beating Chinese pricing while offering superior English-language performance and enterprise compliance features will recapture market share [3]. The immediate question is whether Anthropic and Google will follow with their own cuts. Anthropic's Sonnet 5 introductory pricing expires August 31, creating a natural decision point. DeepSeek's same-day V4 launch suggests Chinese providers intend to compete on capability as well as cost, setting up an extended price war through the second half of 2026 [[2]](https://finance.yahoo.com/technology/ai/articles/openai-cuts-gpt-5-6-173045044.html) [3]. ## Companies mentioned ## Further sources [[1] The Decoder — OpenAI goes full China pricing mode with an 80 percent cut to its… ↗](https://the-decoder.com/openai-goes-full-china-pricing-mode-with-an-80-percent-cut-to-its-most-affordable-gpt-5-6-model/) [[2] Yahoo Finance — OpenAI cuts GPT-5.6 Luna and Terra prices by up to 80% ↗](https://finance.yahoo.com/technology/ai/articles/openai-cuts-gpt-5-6-173045044.html) [[3] 36Kr — GPT-5.6 Sees Sharp Price Cut: Luna Input Pricing Undercuts DeepSeek, Fol… ↗](https://eu.36kr.com/en/p/3919319636290946) [[4] CNBC — OpenAI cuts prices for two of its GPT-5.6 AI models as companies grow se… ↗](https://www.cnbc.com/2026/07/30/open-ai-price-cut-gpt.html) [[5] Basic Tutorials — GPT-5.6: OpenAI Cuts Prices – Luna Becomes 80% Cheaper ↗](https://basic-tutorials.com/news/gpt-5-6-openai-cuts-prices-luna-becomes-80-cheaper/) The stories that matter, in one email. Free — unsubscribe anytime.