Last updated: August 19, 2026 · By Vishal Swami, Founder & Lead AI Reviewer, AISagely
OpenRouter cut GPT-5.6 Sol's price in half on August 17, 2026, and Vercel's AI Gateway matched it the same day. Both discounts run through September 18, 2026, and apply automatically to normal API traffic — you don't need to change a line of code to get the lower rate.
Short answer: GPT-5.6 Sol now costs $2.50 per million input tokens and $15 per million output tokens on OpenRouter and Vercel AI Gateway, down from OpenAI's standard $5/$30, through September 18, 2026. The discount doesn't apply to bring-your-own-key requests or OpenAI's direct API, which still bills the full rate.
Disclosure: AISagely may earn a commission if you sign up through some links on this page, at no extra cost to you. OpenAI itself has no affiliate program tied to this specific promotion — I have no financial stake in which router you pick.
I've been calling GPT-5.6 Sol through the standard API tier since it went broadly available in July, mostly for coding-agent work, so a 50% price swing on the router I sometimes use isn't a hypothetical for me — it changes what I'd actually recommend switching to this month. In my testing this week I pulled the live per-token math off OpenRouter's own model page, then checked the same numbers against OpenAI's official developer docs and Vercel's changelog to see exactly where the discount does and doesn't apply.
How I checked these numbers
I confirmed the discounted rate directly on OpenRouter’s GPT-5.6 Sol pricing page on August 19, 2026, then cross-checked OpenAI's undiscounted, official rate on the OpenAI developer platform’s model card. I confirmed the same promotion is live on Vercel through Vercel’s own changelog post, which is the most specific public source on dates and exclusions. I didn't take OpenRouter's or Vercel's marketing copy at face value — I checked whether the discount actually changes the cheapest way to run Sol, or just makes an already-mediocre option look better for a month.
Top picks at a glance
| Option | Sol price now (in / out per 1M) | Discount applies? | Best for |
|---|---|---|---|
| OpenRouter | $2.50 / $15.00 | Yes, through Sep 18, 2026 | Cheapest metered access right now |
| Vercel AI Gateway | $2.50 / $15.00 | Yes, through Sep 18, 2026 | Teams already billing through Vercel |
| OpenAI direct API (your own key) | $5.00 / $30.00 | No — BYOK is excluded | Full control, no router dependency |
| ChatGPT Plus app | Flat $20/month, no per-token billing | Not applicable | Casual use, no API setup |
| ChatGPT Free app | $0 — Luna only, not Sol | Not applicable | Trying the GPT-5.6 family for free |
Prices move fast in this category — confirm the current rate on OpenRouter’s model page before you commit real spend, since this is a time-limited promotion, not a permanent price cut.
Best overall: OpenRouter
If you're already calling GPT-5.6 Sol through an API key, OpenRouter is the cheapest way to keep doing that right now. The discount isn't limited to the default tier — OpenRouter confirmed it also covers the flex tier (as low as $1.25 input / $7.50 output per million tokens) and priority/fast mode, plus cached tokens and batch API requests. Cache reads run $0.25 per million tokens and cache writes $3.125, both half the standard rate, so a workload that reuses a long system prompt sees the discount stack with caching savings rather than compete with them.
Worth naming the honest catch: this discount only applies to requests billed through OpenRouter's own credits, not to bring-your-own-key traffic, and it's explicitly temporary — it ends September 18, 2026, after which pricing reverts to the standard $5/$30 rate. There's also a skeptical read on why this promotion exists at all: OpenRouter and Vercel are two of the main public data sources analysts use to estimate model market share, and a temporary discount that doubles transaction volume on exactly those platforms can make adoption look stronger than it is on OpenAI's actual paid traffic, which didn't get a price cut. That doesn't make the savings fake — the lower bill is real if you route through OpenRouter during the promo window — but it's a reason not to assume this rate is where Sol settles long-term. Get started at Openrouter.
Best for value: Vercel AI Gateway
Vercel matched OpenRouter's discount the same day, and its version applies across every service tier, region, and token type — including fast mode, cached tokens, cache writes, and long-context requests over 272K input tokens. Vercel charges zero markup on top of the provider's list price, so the discounted $2.50/$15 rate is exactly what you pay, not a router fee on top of it. Like OpenRouter, the discount excludes bring-your-own-key requests, since those bill directly against your own OpenAI account instead of Vercel's.
The real reason to pick Vercel over OpenRouter isn't the price — it's identical during the promo — it's whether you're already inside Vercel's ecosystem. If your app is deployed on Vercel and you're tracking spend through their dashboard anyway, routing GPT-5.6 Sol through AI Gateway means one invoice instead of two accounts to reconcile. If you're not already a Vercel customer, there's no reason to set up an account just for this promotion when OpenRouter offers the same rate with a lower-friction signup.
Best free option: ChatGPT's free plan
There's no genuinely free way to call GPT-5.6 Sol's API — every option above bills per token, discount or not. The closest thing to free is using GPT-5.6 through the ChatGPT app itself. OpenAI's free plan gives unlimited chat access to GPT-5.6 Luna, the smallest model in the family, and free and Go-tier users can reach GPT-5.6 Terra through ChatGPT Work and Codex specifically — but Sol itself, the model this discount is about, requires a Plus subscription ($20/month flat) or higher.
That's a meaningfully different product than the API: no per-token billing, no rate math, just a monthly cap on usage instead. If you want to know whether Sol is even worth paying for before you touch the API at all, my how to use ChatGPT guide covers the free and Plus tiers in more depth, and how to use ChatGPT Codex is the more relevant starting point if coding is why you're looking at Sol pricing in the first place.
How to choose the right option
Start with how you're actually calling the model. If you're hitting the API directly for a side project or a small production workload, OpenRouter's discounted rate is the easiest win available this month — switching a base URL is a smaller lift than most pricing changes require. If your infrastructure already runs through Vercel, use AI Gateway instead and skip managing a second router account for an identical rate.
If you're using your own OpenAI API key rather than routing through a third party, this promotion doesn't help you at all — BYOK bills at the standard $5/$30 rate regardless of what OpenRouter or Vercel are advertising, and there's no workaround short of actually routing traffic through one of them. My managing AI coding costs at scale guide covers the broader habit of tracking spend across model tiers, which matters more once a temporary discount like this one expires. And if Sol's price, even discounted, still feels steep for your workload, it's worth comparing against DeepSeek V4 Flash or checking my wider best AI models roundup before assuming Sol is the only option — my GPT-5.6 vs. Claude Fable 5 comparison is also useful if you're deciding between frontier models rather than just chasing the router with the lowest bill.
My verdict
The discount is real and it's a genuine 50% off, not a marketing exaggeration — I confirmed the same numbers independently on OpenRouter's page, Vercel's changelog, and OpenAI's own developer docs. If you already route GPT-5.6 Sol traffic through either platform, there's no reason not to take the lower rate while it lasts. What I wouldn't do is build a permanent cost model around it: the promotion has a hard end date of September 18, 2026, it excludes BYOK entirely, and there's a real question about whether it's aimed more at inflating usage numbers on the two platforms analysts watch than at making Sol genuinely cheaper for OpenAI's average customer. Use the savings if you're already in this workflow. Don't switch your whole stack to OpenRouter or Vercel on the strength of a month-long promo alone. If cost is your main concern going forward, my piece on AI getting way too expensive is a useful reality check before you lock in any router as a long-term default, and accelerating GPT-5.6 Sol with Ultrafast is worth a read if speed, not price, ends up being your actual bottleneck.
Frequently Asked Questions
What is the new GPT-5.6 Sol price on OpenRouter?
$2.50 per million input tokens and $15 per million output tokens, down from OpenAI's standard $5/$30 rate. Cached input tokens run $0.25 per million and cache writes $3.125 per million, both half the normal price.
How long does the GPT-5.6 Sol discount last?
Through September 18, 2026, on both OpenRouter and Vercel AI Gateway. After that date, pricing is expected to revert to the standard $5/$30 rate unless either platform extends the promotion.
Does the discount apply if I use my own OpenAI API key?
No. Bring-your-own-key (BYOK) requests bill directly against your OpenAI account at the standard rate. The 50% discount only applies to requests billed through OpenRouter's or Vercel's own credits.
Is there a free way to use GPT-5.6 Sol?
Not through the API. The closest free option is ChatGPT's free plan, which gives unlimited access to GPT-5.6 Luna (not Sol) and, for free and Go-tier users, Terra access through ChatGPT Work and Codex. Sol itself requires a $20/month ChatGPT Plus subscription or an API key.
Should I switch from OpenAI's direct API to OpenRouter for this discount?
If your workload is flexible enough to route through a third party, the savings are real for the next month. If you need a direct account relationship with OpenAI, or you're not comfortable with a router sitting between you and the model, it's not worth restructuring your setup for a discount with a fixed end date.