# GPT-5.6 Sol promo price ends Nov 21: what to lock in now

URL: https://infro.io/blog/gpt-5-6-sol-promo-price-ends-nov-21-what-to-lock-in-now
Author: INFRO Editorial (https://infro.io/blog/editorial)
Publisher: INFRO
Published: 2026-11-02
Last modified: 2026-11-02
Last reviewed: 2026-11-02 (next review due 2026-12-02)
Topic: Cost and FinOps
Format: Brief

> OpenAI guarantees GPT-5.6 Sol's promotional rate at least through 21 November 2026 and publishes no rate after it [^ai-updates]. Budget the post-21-November rate as unknown, model a range, and choose per workload: commit, switch tier or cap.

Key takeaways:
- OpenAI guarantees GPT-5.6 Sol's promotional rate at least through 21 November 2026; that is a floor, not a stated end, and no later rate is published.
- Plan as if the step lands on 22 November, and model a range of step sizes instead of one invented number.
- Two of three typical workloads are about two-thirds output cost, so an output-weighted increase hits them hardest.
- Do not size a spend commitment on the promotional run-rate; a straight-line forecast does not know about price steps.
- Per workload, choose commit, switch tier (after your own eval) or cap, and set the cap before the date.

## What is guaranteed, and what is not

OpenAI guarantees the GPT-5.6 Sol promotional rate at least through 21 November 2026, and INFRO's updates timeline advises that budgets running past that date should use the standard rate [1]. OpenAI's wording is "available at least through November 21, 2026": a floor, not an end date, and no later rate is dated [2].

Three things this article could not confirm, because OpenAI's own pricing page was not readable on 2 November:

- **The post-promotion rate.** None is published in the sources read.
- **Which rate is the promotional one.** INFRO's catalog lists GPT-5.6 Sol at $4.00 input, $20 output and $0.40 cached input per million tokens, checked 5 October [3]. Whether that is the promotion or the standard rate is not stated there.
- **What the promotion covers.** Cached input, Batch, Flex, Fast mode and long context may or may not follow the same step. Read OpenAI's page before you sign anything.

So a defensible forecast has two columns: the listed rate until 21 November, and a range after it.

## Before and after on three workloads

> **Example: Illustrative volumes and step sizes**
> The volumes and the 1.5x and 2x steps are invented to show the arithmetic. The step is applied to input, cached and output rates alike. "Now" uses the $4 / $0.40 / $20 rates above [3]; the GPT-6 Sol column uses OpenAI's list standard rates of $2 input, $0.20 cached and $10 output [4].
> 
> - **A. Support assistant:** 1M requests a month, each 500 fresh input, 1,500 cached input, 300 output tokens.
> - **B. Nightly summaries:** 300,000 records a month, each 4,000 input and 400 output tokens, nobody waiting.
> - **C. Agent tasks:** 20,000 tasks a month, each 60,000 input (45,000 cached) and 8,000 output tokens, reasoning included.
> 
> | Workload | Now, per month | If step is 1.5x | If step is 2x | GPT-6 Sol list |
> | --- | --- | --- | --- | --- |
> | A. Support assistant | $8,600 | $12,900 | $17,200 | $4,300 |
> | B. Nightly summaries | $7,200 | $10,800 | $14,400 | $3,600 |
> | C. Agent tasks | $4,760 | $7,140 | $9,520 | $2,380 |
> 
> Table: Monthly provider cost on GPT-5.6 Sol before and after a hypothetical step, against GPT-6 Sol at list.

Workloads A and C are about 70% and 67% output cost, so a step that falls mainly on output would hit them harder than these uniform figures show. In November itself, a step on 22 November puts 9 of 30 days at the new rate: workload A at 2x lands near $11,180 instead of $8,600.

Workload B has a second lever. OpenAI lists Batch and Flex at half the standard rate [4]; if Batch is offered for GPT-5.6 Sol on your account, B at today's rate halves to $3,600, which is below its 2x figure of $14,400 by a wide margin. Confirm the model is listed for Batch first.

## Decision list: commit, switch tier or cap

Make the call per workload, not per account.

1. **Commit** only where the workload is stable, you have already passed it on GPT-5.6 Sol, and the contract fixes a dollar rate for the term. Do not size a commitment on the promotional run-rate. INFRO's commitment forecast is the amount consumed plus the trailing 28-day daily mean times the days remaining, and it states that it does not model price changes [5]. Enter the step yourself and compare.
2. **Switch tier** where GPT-6 Sol could pass your own eval. At list it is half the GPT-5.6 Sol input and output rate in the table, and OpenAI lists GPT-6.1 Sol with the same rates and a lower cached-input rate [4]. The saving only counts if the pass rate holds, so measure cost per successful task and re-run on the new snapshot; the method is in [GPT-6 Astra, Sol or Luna](https://infro.io/blog/gpt-6-astra-sol-luna-tiers).
3. **Cap** everything that stays, before 21 November. A blocking budget admits a request only if its worst-case estimated cost fits the room left, and without `max_tokens` the worst case assumes the model's maximum output [6]. Set `max_tokens` on workload C in particular, since its output is reasoning-heavy.

> **Warning: Pins and aliases**
> An application pinned to GPT-5.6 Sol keeps paying whatever it costs after the step. The pin is a decision; make it on the post-step rate.

## Steps to take this week

- [ ] Pull November-to-date cost per project for GPT-5.6 Sol, split into fresh input, cached input and output.
- [ ] Read OpenAI's current pricing page and record the rate, the date and what the promotion covers.
- [ ] Put a 22 November row in your price schedule, as in [the introductory-pricing guide](https://infro.io/blog/ai-model-launch-pricing-budgets), and forecast December at two or three step sizes.
- [ ] Ask your account contact whether your rate is fixed in dollars or relative to list.
- [ ] Start the GPT-6 Sol eval on the highest-spend workload first.
- [ ] Set blocking budgets and `max_tokens` on the rest.

If requests go through INFRO, per-request estimated cost by project and key shows which workload dominates, and entering your contract rate under custom rates makes estimates use it ahead of list. Budgets and commitment tracking work on your own OpenAI account; OpenAI keeps billing you directly for the inference.

## Questions

### Does the promotional price end on 21 November?

Not necessarily. OpenAI's wording is that promotional pricing is available at least through 21 November 2026, so that date is the earliest the promotion can end, not a stated end [2]. Budget as if it ends then and treat anything later as upside.

### Will my negotiated rate cover the step?

Unknown until you read the contract. Ask whether a negotiated rate for GPT-5.6 Sol is fixed in dollars or defined relative to list, because a relative rate moves when list moves.

## Sources

1. AI updates: releases, price changes and shutdowns (entry: GPT-5.6 Sol promotional price guaranteed until 21 Nov) — INFRO. https://infro.io/ai-updates (accessed 2026-11-02)
2. Introductory AI model prices: how to budget for the step-up — INFRO. https://infro.io/blog/ai-model-launch-pricing-budgets (accessed 2026-11-02)
3. GPT-5.6 Sol API pricing — INFRO. https://infro.io/models/openai/gpt-5.6-sol (accessed 2026-11-02)
4. OpenAI API pricing in 2026: GPT-6 tiers, Batch, Flex and Fast — INFRO. https://infro.io/blog/openai-api-pricing-2026 (accessed 2026-11-02)
5. Commitments — INFRO Docs. https://infro.io/docs/cost/commitments (accessed 2026-11-02)
6. Spend controls — INFRO Docs. https://infro.io/docs/org/spend-controls (accessed 2026-11-02)