OpenRouter halved the price of OpenAI's flagship GPT-5.6 Sol — $2.50 per million input tokens and $15 per million output, down from $5/$30 — automatically and with no code change, in a limited promotion that Vercel AI Gateway matched over the same window. OpenAI's own API price is unchanged.
2026-08-18 · by Moe Ameen
OpenRouter — the multi-model API marketplace that routes one OpenAI-compatible endpoint to hundreds of models — applied a 50% discount to OpenAI's flagship GPT-5.6 Sol in mid-August 2026, dropping the standard rate of $5 per million input tokens and $30 per million output to $2.50 and $15. The cut takes no code change: requests to `openai/gpt-5.6-sol` bill at the lower rate automatically. Vercel AI Gateway applied the identical discount over the same window — its priority tier fell from $10/$60 to $5/$30 — and the reduction covers the batch, flex, and cached token paths as well.
The discount is limited to those two third-party gateways. OpenAI's own first-party API price for Sol is unchanged, so a developer calling OpenAI directly still pays the full rate — where you route the request now decides whether you get the promotion. GPT-5.6 Sol itself is OpenAI's top GPT-5.6 tier, a frontier reasoning, writing, and tool-orchestration model with a 1M-token context window, and on OpenRouter it is served across several providers (OpenAI, Azure, Amazon Bedrock) with routing modes for balanced cost and speed, fastest throughput, or highest tool-calling accuracy.
The promotion drew scrutiny almost immediately. Research firm SemiAnalysis pointed out that OpenRouter and Vercel account for a negligible share of OpenAI's total inference volume, yet they are among the primary public data sources the industry uses to estimate model market share — so a half-price promo that could double Sol's transaction volume on those platforms may inflate its apparent share against rivals like Anthropic and DeepSeek. Treat the dates as the promotional window OpenRouter published (announced mid-August, running through September 18); confirm the live rate on the model page before you budget around it.
If you run your own generation stack on OpenRouter, the practical move here is small and time-boxed: through September 18 your GPT-5.6 Sol text calls — scripts, blog drafts, newsletter copy — cost half as much, so it is a reasonable moment to front-load batch work you were going to do anyway. But keep the SemiAnalysis subtext in view: the token was never the expensive part of shipping content. Drafting 900 words with Sol costs cents; turning those 900 words into a captioned vertical video, a brand-exact carousel, a quote card, a blog post, and a newsletter — then publishing all of it on a schedule across every platform — is the actual work, and no per-token discount touches it.
That downstream half is what [Kompozy](/) owns. Point it at one source and it generates the full spread — [Persona Shorts](/glossary/persona-shorts) and face-locked avatar video, [Carousel Posts](/glossary/hyperframes), Quote Graphics, Blog Articles, and Email Newsletters — every piece written in one voice through your [Persona Brief](/glossary/persona-brief), then reframed per platform and pushed to the eight social platforms plus blog and email from a single queue on [Autopilot](/glossary/autopilot). Kompozy supports bring-your-own-key on its Founding tier, so the model bench stays your choice while the engine adds everything the model can't do alone. The half-price Sol tokens are a pleasant tailwind for a month; the system that turns any model's output into a published content week is the part that still matters after the promo ends.
It is 50% off standard — $2.50 per million input tokens and $15 per million output, down from $5/$30 — applied automatically to openai/gpt-5.6-sol with no code change. The same cut hit the priority tier ($10/$60 down to $5/$30) and the batch, flex, and cached paths.
OpenRouter published the promotion as running through September 18, 2026 (announced in mid-August). After that the rate is set to return to standard, so confirm the live price on the model page before you budget around it.
No. Only OpenRouter and Vercel AI Gateway cut the price. OpenAI's first-party API bills Sol at the unchanged standard rate, so where you route the request determines whether you get the lower price.
SemiAnalysis argued that OpenRouter and Vercel are a tiny share of OpenAI's inference volume but are primary public sources for market-share estimates, so a half-price promo that boosts Sol's transactions there could inflate its apparent share against competitors like Anthropic and DeepSeek.