# OpenAI cuts GPT-5.6 Sol output pricing by one-third for three months > Source: > Published: 2026-08-21 20:14:43+00:00 # OpenAI cuts GPT-5.6 Sol output pricing by one-third for three months **The flagship model now costs $4 per million input tokens and $20 per million output tokens, while paid-plan usage limits stay put.** By [Ryan Merket](/author/ryan-merket) ยท Published Primary source: [OpenAI on X](https://x.com/OpenAI/status/2090885187634905500) ## Why it matters The discount cuts output-heavy GPT-5.6 Sol workloads by up to one-third, while its temporary structure lets OpenAI lower adoption costs without committing to a permanent price reset. [OpenAI said on August 21st](https://x.com/OpenAI/status/2090885187634905500) that it cut API and purchased-credit pricing for [GPT-5.6 Sol](/models/openai/gpt-5.6-sol) through at least November 21st, giving developers and agent users three months of lower inference costs without changing the usage included in ChatGPT subscriptions. [https://x.com/OpenAI/status/2090885187634905500](https://x.com/OpenAI/status/2090885187634905500) The discount is larger for generated tokens, usually the expensive side of a reasoning workload. [GPT-5.6 Sol API pricing](https://developers.openai.com/api/docs/models/gpt-5.6-sol) fell from $5 to $4 per million input tokens, a 20% reduction. Cached input dropped from $0.50 to $0.40 per million tokens. Output fell from $30 to $20 per million tokens, cutting that portion of the bill by one-third. A workload consuming one million input tokens and producing one million output tokens would therefore cost $24 under the promotion, down from $35. That is a 31% reduction before tool-call fees, priority processing, cache writes, or the higher rates applied to prompts containing more than 272,000 input tokens. The actual savings depend on the workload's mix of input, cached context, and generated output. The same proportions apply to purchased credits used for eligible ChatGPT Work and Codex activity. OpenAI's updated [credit rate card](https://help.openai.com/en/articles/11481834) prices one million GPT-5.6 Sol input tokens at 100 credits, cached input at 10 credits, and output at 500 credits. The previous rates were 125, 12.5, and 750 credits, respectively. That distinction matters for customers reading the announcement as a broader expansion of their plans. OpenAI said included usage for Plus, Pro, and Business subscriptions remains unchanged. Its rate card also says the promotion applies to eligible activity paid for with purchased credits, while included plan usage, five-hour and weekly limits, and legacy credit rates remain intact. For Plus and Pro users, credits act as a pay-as-you-go extension after included allowances run out. OpenAI's [credit documentation](https://help.openai.com/en/articles/12642688-using-credits-for-flexible-usage-in-chatgpt-free-go-plus-pro) says supported features can draw from a shared balance across Codex, ChatGPT Work, and ChatGPT for Excel, depending on account eligibility. Business and enterprise arrangements vary with workspace billing and credit settings. ### A temporary cut across OpenAI's flagship tier OpenAI launched the [GPT-5.6 family](https://openai.com/index/gpt-5-6/) on July 9th with three tiers: Sol as the flagship model, Terra as the mid-priced option, and Luna for high-volume work. Sol originally entered general availability at $5 per million input tokens and $30 per million output tokens. OpenAI then reduced Luna's price by 80% and Terra's by 20% on July 30th. The Sol promotion extends that repricing across the full GPT-5.6 lineup six weeks after general availability, while preserving a meaningful gap between the three tiers. Terra remains priced below Sol at $2.50 per million input tokens and $15 per million output tokens, and Luna costs $1 and $6, respectively. Sol carries the `gpt-5.6` API alias and supports a 1.05 million-token context window, up to 128,000 output tokens, tool use, computer use, hosted shell access, and reasoning settings from none through max. Those capabilities can produce large token bills in long-running coding and research agents, particularly when a task generates substantial output or coordinates repeated tool calls. The new rates make Sol easier to test against cheaper GPT-5.6 variants without permanently lowering its listed position in OpenAI's lineup. Output-heavy agents receive the largest benefit, while applications dominated by input or cached context get a 20% reduction. OpenAI retains the option to restore the original rates after November 21st. The promotion is already available through the API and is rolling out to eligible purchased-credit usage in ChatGPT Work and Codex. Standard ChatGPT conversations and their existing subscription allowances are outside the price change.