GPT-5.6 Sol Promo Pricing Ends November 30: What Your API Bill Does Next
GPT-5.6 Sol's discounted API price of $4 per million input tokens and $20 per million output tokens is promotional, and OpenAI's model page now says it is available until November 30, 2026. The earlier list price was $5 input and $30 output, so a workload that sends and receives equal token volumes would cost about 46% more if the rate reverts. That affects any team that budgeted Sol at the August 21 cut.
Key facts
- August 21, 2026: OpenAI cut Sol from $5 to $4 input and from $30 to $20 output per million tokens, with cached input moving from $0.50 to $0.40.
- OpenAI's GPT-5.6 model page states: "GPT-5.6 Sol's promotional pricing is available until November 30, 2026."
- The AWS announcement for Amazon Bedrock, also dated August 21, says the reduced price is available "at least through November 21, 2026."
- Terra ($2 input, $12 output) and Luna ($0.20 input, $1.20 output) were cut on July 30 and are described as permanent reductions.
- Prompts above 272K input tokens are billed at 2x the input rate and 1.5x the output rate for the whole request.
- The OpenAI page does not state what Sol costs after the promotion ends. The pre-cut list price was $5 and $30.
What the end date is
Two vendor sources give two dates. OpenAI's own page says November 30. The Bedrock notice says "at least" November 21, which is a floor and not an end date. Reading both together, the safe planning date is November 21 for anything running on Bedrock and November 30 for OpenAI's own API, with no published price for the day after. Today is November 6, so there are 24 days left on the later date.
The promotion covers more than the standard text rate. The August 21 announcement says it also applies to Fast mode, long-context requests, and Batch and Flex processing. It reaches API usage and purchased credits. ChatGPT subscription usage is described as unchanged, so subscribers see no difference either way.
What a reversion costs on real workloads
The figures below compare the $4/$20 promo rate with the earlier $5/$30 list price, for a given volume of input and output tokens. Cached input is excluded to keep the arithmetic visible.
- Input-heavy, 10M input and 1M output: $60 at promo, $80 at list. That is 33% more.
- Balanced, 1M input and 1M output: $24 at promo, $35 at list. That is 46% more.
- Output-heavy, 1M input and 5M output: $104 at promo, $155 at list. That is 49% more.
The more your product generates, the more a reversion hurts, because the output rate moves by 50% ($20 to $30) while the input rate moves by 25% ($4 to $5). Agents, code generation and long-form drafting sit at the expensive end. Classification and retrieval-style calls with short answers sit at the cheap end.
Where Sol sits against its alternatives
At $4/$20, Sol is cheaper than Claude Opus 5, which a Yellow.com report on the cut lists at $5 input and $25 output. At $5/$30 the order flips. Within OpenAI's own line, Terra at $2/$12 costs half the Sol promo rate on input and 60% of it on output, and that price is described as permanent. Any comparison you ran in September or October used the promotional number, so rerun it with list pricing before you commit to a quarter.
OpenAI has repriced this lineup repeatedly. Our earlier look at the GPT-5.6 Luna price cut argued that launch-week and promotional rates are provisional. Google has set a similar date for its own flagship: see Gemini 4 Argon pricing doubling on December 1.
What to do before November 30
- Pull your last 30 days of Sol usage and split it into input, cached input and output tokens. Multiply by $5, $0.50 and $30 per million to get the worst-case monthly bill.
- Set a budget alert at that worst-case figure for December 1, and mark November 21 as an earlier checkpoint if you call Sol through Bedrock.
- Test your three highest-volume Sol prompts on Terra. If quality holds, the saving is permanent and does not depend on a promotion.
- Check for long-context calls above 272K input tokens. They pay the 2x and 1.5x multipliers on top of whichever base rate applies.
- Re-read OpenAI's GPT-5.6 model page in the last week of November, since the end date differs between sources.
For more pricing coverage, browse the full article archive.
Questions people ask
When does GPT-5.6 Sol's discounted price end?
OpenAI's model page says promotional pricing is available until November 30, 2026. The AWS Bedrock announcement says at least through November 21, 2026.
How much will GPT-5.6 Sol cost after the promotion?
OpenAI's page does not say. The price before the August 21 cut was $5 per million input tokens and $30 per million output tokens, so that is the figure to budget against.
Do ChatGPT subscribers pay more when the promotion ends?
No change is announced for Pro, Plus or Business subscriptions. The Sol discount applies to API usage and purchased credits.
Are the Terra and Luna price cuts also temporary?
No. OpenAI describes the July 30 reductions to Terra and Luna as permanent.
Sources