OpenAI Cut Its Flagship Model to Half Price, and the Reason Is Sitting on a Leaderboard
GPT-5.6 Sol is 50 percent off through 18 September on OpenRouter and Vercel's gateway, taking it to 2.50 dollars per million input tokens. It follows cuts to Luna and Terra as Chinese models gain ground.
OpenAI is running a 50 percent discount on GPT-5.6 Sol, its flagship model, through 18 September. On OpenRouter the listed price falls from 5 dollars to 2.50 per million input tokens and from 30 to 15 dollars per million output tokens. The discount extends to the batch, flex and priority tiers, which takes flex down to as low as 1.25 dollars in and 7.50 out. Vercel is running the same promotion on its AI Gateway. It applies automatically on the OpenAI provider and does not apply if you bring your own API key.
A token, if the word is new to you, is roughly a chunk of a word: a million tokens is somewhere around 750,000 words of English. So the practical effect is that feeding a model the equivalent of several long novels now costs a couple of dollars rather than five.
This is not an isolated gesture. OpenAI has already cut Luna by 80 percent and Terra by 20 percent, and the pattern points in one direction. On OpenRouter, which routes developer requests across hundreds of models and therefore acts as a rough scoreboard for what people actually choose, Chinese models have been taking share. DeepSeek and Kimi both compete near the top on capability while pricing aggressively, and developers picking a model through a gateway compare cost per token on the same screen as benchmark scores. When your competitor is a quarter of your price and close enough on quality, a price cut is the fastest lever you have.
Worth knowing before you get excited: this is a promotion with an end date, not a new list price. Some analysts have pointed out that discounting specifically on third-party gateways is also a good way to move visible market-share numbers, since gateway rankings are widely quoted. Both things can be true at once. And the discount does not touch OpenAI’s own direct API pricing, so where you buy matters.
What this means for you: if you use ChatGPT rather than the API, nothing changes, because consumer subscriptions are priced separately. If you build things, this is a genuinely good month to run the experiment you have been putting off: batch jobs, large document processing, anything where token cost was the reason you said no. Just do not architect around the price. Build so you can swap models, because the whole point of this story is that model prices are now moving fast in both directions. The broader trend is the reassuring part: competition, including from open-weight Chinese labs, is pushing the cost of good AI down for everyone.
Sources
Your Next PC Upgrade Just Got Five Times More Expensive, and AI Is the Reason
DDR5 memory prices are up nearly 500 percent in twelve months. A 128GB kit that once bottomed out at 329 dollars now lists at 3,399. The cause is not scalping, it is AI data centres eating the wafers.
