OpenAI’s price cuts highlight convergence of open and closed AI models

OpenAI cut the price of its GPT-5.6 Luna model by 80% on July 30, lowering costs to $0.20 per million input tokens and $1.20 per million output tokens, while the larger GPT-5.6 Terra received a 20% reduction to $2 per million input tokens and $12 per million output tokens. The cuts, driven by optimizations from the sibling GPT-5.6 Soul model, reflect growing competitive pressure from Anthropic and Chinese AI labs, and highlight the shrinking cost gap between proprietary and open-weight models like Moonshot AI's Kimi K3, which is priced at $3 per million input tokens and $15 per million output tokens.

Via gizmodo.com OpenAI’s price cuts highlight convergence of open and closed AI models An 80% price slash on GPT-5.6 Luna signals that the moat between proprietary and open-weight AI is shrinking fast OpenAI just took a machete to its pricing. On July 30, the company announced an 80% reduction on its GPT-5.6 Luna model, bringing costs down to $0.20 per million input tokens and $1.20 per million output tokens. The larger GPT-5.6 Terra got a 20% haircut, landing at $2 per million input tokens and $12 per million output tokens. The efficiency engine behind the cuts OpenAI attributed the reductions to internal optimizations powered by GPT-5.6 Soul, a sibling model in the same family. Those optimizations, including kernel rewrites and speculative decoding, reportedly delivered 20-35% efficiency gains in inference workloads. The timing is not coincidental. Back in June 2026, reports surfaced that OpenAI was exploring drastic token price reductions as a defensive measure against mounting competitive pressure, particularly from Anthropic and a wave of Chinese AI labs that have been aggressively undercutting Western providers on cost. Open-weight models are no longer the budget option Moonshot AI’s Kimi K3, an open-weight model with 2.8 trillion parameters, launched around the same time as OpenAI’s pricing changes. It’s priced at $3 per million input tokens and $15 per million output tokens when accessed through hosted APIs. That puts it above OpenAI’s newly discounted Luna pricing, and roughly in the same neighborhood as Terra. Running a 2.8-trillion-parameter open model still requires serious infrastructure investment for self-hosting. The GPU clusters, the memory requirements, the engineering talent to keep everything humming: none of that is cheap. Enterprises are shopping smarter This convergence is partly driven by how enterprise customers actually use AI in production. Increasingly, companies deploy model routers, systems that automatically route queries to the cheapest model capable of handling the task. A simple classification request might go to a lightweight open model. A complex reasoning task gets sent to a frontier model. That behavior puts relentless downward pressure on pricing. US labs are feeling this acutely, as enterprises treat raw intelligence as a commodity and optimize ruthlessly for cost per quality unit. Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy https://cryptobriefing.com/editorial-policy/ .