How much is GPT-5.6 Sol on Vercel AI Gateway during the discount?

Default is $2.50 per million input tokens and $15.00 output; Flex is $1.25 and $7.50; Priority or fast mode is $5.00 and $30.00. The discount runs through September 18, 2026.

What does the discount cover?

It applies across every service tier including fast mode, and covers cached tokens, cache writes, long-context requests and all regions.

Should I migrate workloads to take advantage?

Testing, yes. Migrating, carefully — the rate doubles on September 19, so anything whose economics only work at the discounted price is not viable at the standard one. Price your real volume at both rates before committing.

Vercel Halves GPT-5.6 Sol for One Month

A shopfront roller shutter half lowered at dusk, warm light spilling from the narrowing gap — PricingAugust 17, 2026·2 min read·By the Automation Squad Research
Vercel Halves GPT-5.6 Sol for One Month
A month-long discount is a good test window and a bad thing to build a budget on.

By Robert MacKelfresh

Founder, Automation Squad · August 17, 2026

The short answer

Vercel discounted openai/gpt-5.6-sol by 50% on AI Gateway through September 18, 2026. Default pricing becomes $2.50 per million input tokens and $15.00 output, Flex $1.25 and $7.50, and Priority or fast mode $5.00 and $30.00. It covers cached tokens, cache writes, long-context requests and all regions. BYOK requests are excluded and bill at your own provider rates.

Price tableDeadline: September 18, 2026

The discounted tiers, what's excluded, and how to use a one-month window

Straightforward pricing, with one exclusion that catches people and one date that matters more than the rates.

Check whether you are on BYOK, because that is the whole catch
Bring-your-own-key requests run against your own provider account and bill at whatever rate you have with them. The discount does not touch them. If you moved to BYOK for control, you get none of this.
Use the month to test, not to save
A discount that ends on a fixed date is a good reason to trial a model on real work and a bad reason to change your architecture. Anything you migrate this month gets 50% more expensive on September 19.
Look at Flex if you have work nobody is waiting on
At $1.25 and $7.50 it is the cheapest tier by a wide margin. Batch and background jobs generally do not need Priority, and plenty of teams are on Default out of habit rather than requirement.
Model your real cost at the post-discount rate before deciding anything
Take your expected monthly volume and price it at both the discounted and the standard rate. If the decision only works at the discounted number, it is not a decision, it is a promotion.
Diary September 18
One-month discounts end quietly. Put a reminder in for a few days before so the increase is a decision rather than a surprise on an invoice.

Tier	Input / output per million
Default	$2.50 / $15.00
Flex	$1.25 / $7.50
Priority (fast mode)	$5.00 / $30.00
Covers	Cached tokens, cache writes, long context, all regions
Excluded	BYOK — bills at your own provider rate
Ends	September 18, 2026

Take it with you

VERCEL AI GATEWAY — GPT-5.6 SOL 50% OFF
Model ID: openai/gpt-5.6-sol · ends Sept 18, 2026
Source: https://automationsquad.com/news/vercel-gpt-5-6-sol-discount/

DISCOUNTED PRICES, PER MILLION TOKENS
  Default ................ $2.50 in / $15.00 out
  Flex ................... $1.25 in / $7.50 out    <- cheapest by far
  Priority (fast mode) ... $5.00 in / $30.00 out

COVERS
  Cached tokens · cache writes · long-context requests · all regions
  All service tiers and modes

EXCLUDED — the catch
  BYOK requests run on YOUR provider account and bill at YOUR rate.
  The discount does not touch them.

CHECKLIST
[ ] Am I on BYOK? If yes, none of this applies to me.
[ ] Use the month to TEST, not to migrate.
    Anything I move this month gets 50% more expensive on Sept 19.
[ ] Do I have work nobody is waiting on? -> Flex, not Default
    (a lot of teams sit on Default out of habit)
[ ] Price my real monthly volume at BOTH rates.
    If it only works at the discounted number, it's a promotion,
    not a decision.
[ ] Calendar reminder: Sept 18

Vercel announced on August 17, 2026 that GPT-5.6 Sol is discounted 50% on AI Gateway for the following month, through September 18.

The facts: the model ID is openai/gpt-5.6-sol. Discounted default pricing is $2.50 per million input tokens and $15.00 per million output; Flex is $1.25 and $7.50; Priority, which covers fast mode, is $5.00 and $30.00. Vercel states the discount applies across all service tiers and modes and covers cached tokens, cache writes, long-context requests and different regions. BYOK requests are excluded — those run on your own provider accounts and bill at whatever rate you have with that provider.

Automation Squad's take: promotional pricing is worth exactly one thing, which is a cheap window to find out whether a model is good enough for a job you have been putting off testing. It is worth close to nothing as a basis for an architecture decision, because the rate doubles on a date that is already in the calendar. The tier spread is the more durable observation here — Flex at $1.25 and $7.50 against Priority at $5.00 and $30.00 is a fourfold difference for the same model, and a lot of workloads sit on a faster tier than they need because nobody revisited the default after the first week. The BYOK exclusion is the detail that will catch people, and it catches the more sophisticated users specifically, since bringing your own key is what you do once you care about control.

Run this now: check whether your traffic is BYOK, because if it is, none of this applies to you and you can stop reading. If it is not, pick one workload you have been meaning to trial on GPT-5.6 Sol and run it this month at half price. Separately, look at whether anything currently on Default or Priority could run on Flex — that is a permanent saving rather than a temporary one. Then put September 18 in the calendar, because one-month discounts end quietly and the first sign is usually an invoice.

Questions people are asking

How much is GPT-5.6 Sol on Vercel AI Gateway during the discount?: Default is $2.50 per million input tokens and $15.00 output; Flex is $1.25 and $7.50; Priority or fast mode is $5.00 and $30.00. The discount runs through September 18, 2026.
What does the discount cover?: It applies across every service tier including fast mode, and covers cached tokens, cache writes, long-context requests and all regions.
Are BYOK requests discounted?: No. Bring-your-own-key requests run on your own provider accounts and bill at whatever rate you have with that provider, so the Gateway discount does not apply.
Should I migrate workloads to take advantage?: Testing, yes. Migrating, carefully — the rate doubles on September 19, so anything whose economics only work at the discounted price is not viable at the standard one. Price your real volume at both rates before committing.