AI Tools
GPT-5.6 Sol OpenRouter pricing: what does the model cost after the 50% cut?
GPT-5.6 Sol OpenRouter pricing sits at $2 per million input tokens and $10 per million output tokens as of August 26, 2026, exactly half of OpenAI's direct $4/$20 API rate. Azure and Amazon Bedrock still list the pre-cut $5/$30 and $5.50/$33.
What matters
- OpenRouter's listed price for gpt-5.6-sol is $2.00 per million input tokens and $10.00 per million output tokens, tagged 50% off, verified on the OpenRouter model page August 26, 2026.
- The new resale rate is exactly half of OpenAI's direct promotional price ($4/$20), which OpenAI guarantees at least through November 21, 2026.
- Azure and Amazon Bedrock have not matched the cut: OpenRouter's provider table still shows $5.00/$30.00 for Azure and $5.50/$33.00 for Bedrock as of August 26.
- OpenRouter's $2/$10 equals OpenAI's own Batch tier for the same model, so real-time resale now matches the direct API's deferred-latency price.
What is GPT-5.6 Sol?
GPT-5.6 Sol is OpenAI's flagship frontier model and the most expensive member of the GPT-5.6 family, which also includes GPT-5.6 Terra and GPT-5.6 Luna. OpenAI's model documentation lists a 1,050,000-token context window, 128,000 max output tokens, and a February 16, 2026 knowledge cutoff; OpenRouter lists the release date as July 9, 2026. The gpt-5.6 alias routes requests to GPT-5.6 Sol.
Buyers pay per token, and the model's pricing changed twice in August 2026: OpenAI cut its direct API rate on August 21, and OpenRouter re-cut its resale rate this week. The same model name now carries three different price tags depending on where you call it.
The OpenRouter 50% cut, verified
OpenRouter's model page for openai/gpt-5.6-sol shows a 50% off badge and a headline price of $2 per million input tokens and $10 per million output tokens, verified August 26, 2026. The provider table lists the OpenAI row with two figures per column: $4.00 and $2.00 for input, $20.00 and $10.00 for output, $0.40 and $0.20 for cache read. The first number is OpenAI's list price, the second is the discounted rate OpenRouter charges.
This is OpenRouter's second cut for the model. The same page listed $2.50/$15.00 on August 19, 2026, exactly half of the old $5/$30 direct rate, and this site's August 25 price-cut article carried that verification. After OpenAI's August 21 cut to $4/$20, OpenRouter moved to exactly half of the new rate. The result: OpenRouter now sells real-time access at $2/$10, the same per-token price as OpenAI's own Batch tier on the direct API.
Where OpenAI's cut did not land: Azure and Bedrock
OpenRouter routes gpt-5.6-sol requests to whichever provider you select, and its provider table exposes what each host charges. As of August 26, 2026, Azure still lists $5.00 input and $30.00 output per million tokens, Amazon Bedrock (US) lists $5.50 and $33.00, and the Azure (EU) and Azure (US) rows match Bedrock at $5.50/$33.00. None of those rows carries the 50% off badge.
The practical meaning: OpenAI's August 21 reduction is a direct-API discount only. Teams buying GPT-5.6 Sol through Azure or Bedrock are still on the pre-cut rate, and OpenRouter's provider table is the one place the whole spread is visible in a single view.
One caveat: these are the rates OpenRouter displays for those hosts. Azure and Bedrock marketplace invoices can differ from a reseller's table, so teams on those clouds should confirm their own contract rates before assuming the higher price applies.
What you actually pay on each path
Direct OpenAI API: $4.00 input, $0.40 cached input, $5.00 cache writes, and $20.00 output per million tokens on the standard short-context tier. OpenAI marks the rate promotional, available at least through November 21, 2026, and prompts above 272K tokens are billed at 2x input and 1.5x output for the full request, per the model documentation.
OpenRouter: $2.00 input, $0.20 cache read, and $10.00 output per million tokens. Routing modes choose the provider for you (Balanced, Nitro for speed, Exacto for tool-calling accuracy), and the OpenAI provider row shows 4.37s P50 latency. OpenRouter's price-history panel reports a weighted average price actually paid of $0.8339 per million input tokens and $15.32 per million output tokens, because caching and discounts pull real spend below the listed rate.
Azure and Bedrock via OpenRouter: $5.00/$30.00 and $5.50/$33.00, no promotional tag.
The honest catch: $2/$10 is OpenAI's Batch-tier price, which on the direct API buys deferred latency. OpenRouter sells the same model at that price with real-time routing, but a reseller rate can change without the notice OpenAI's November 21 wording gives you.
Who should buy where
- Price-sensitive workloads: OpenRouter at $2/$10 halves the token bill versus direct, and the $0.20 cache read halves cached-input cost again.
- First-party toolchain and support: buy direct at $4/$20 for OpenAI's usage dashboards, service tiers, and support, with the promo rate locked at least through November 21, 2026.
- Teams already on Azure or Bedrock: confirm your contract rate. OpenRouter's table still shows the pre-cut $5.00/$30.00 and $5.50/$33.00 as of August 26, so moving to its OpenAI row is a 60% discount on input tokens today.
- Anyone routing through OpenRouter: the rate moved twice this month, from $2.50/$15.00 to $2/$10, and OpenRouter publishes no guarantee window like OpenAI's. Budget for reseller price risk.
All figures are short-context standard rates per 1 million tokens, verified August 26, 2026. OpenRouter's provider table does not break out a long-context column, so treat these as short-context rates.
At a glance
| Provider | Input /1M | Cache read /1M | Output /1M | Notes |
|---|---|---|---|---|
| OpenAI direct API | $4.00 | $0.40 | $20.00 | Promotional rate through at least Nov 21, 2026 |
| OpenRouter (OpenAI) | $2.00 | $0.20 | $10.00 | 50% off tag; matches OpenAI Batch tier |
| Azure | $5.00 | $0.50 | $30.00 | Pre-cut rate, per OpenRouter provider table |
| Amazon Bedrock (US) | $5.50 | $0.55 | $33.00 | Pre-cut rate, per OpenRouter provider table |
FAQ
Is GPT-5.6 Sol cheaper on OpenRouter than on OpenAI's API?
Yes. OpenRouter lists gpt-5.6-sol at $2 per million input tokens and $10 per million output tokens as of August 26, 2026, exactly half of OpenAI's direct promotional rate of $4/$20. OpenAI's rate holds at least through November 21, 2026; OpenRouter publishes no equivalent guarantee.
Why does OpenRouter show $4 and $2 for the same model?
The provider table shows the OpenAI row with two numbers: the $4.00 list price OpenAI posts and the $2.00 discounted rate OpenRouter charges, tagged 50% off. Azure, Bedrock, and the Azure EU/US rows show a single pre-cut rate each.
Did Azure and Amazon Bedrock lower GPT-5.6 Sol prices too?
Not as of August 26, 2026. OpenRouter's provider table still lists Azure at $5.00/$30.00 and Amazon Bedrock (US) at $5.50/$33.00 per million tokens, the pre-cut rates. OpenAI's August 21 reduction applies to the direct API only, per its changelog.
Related reading
GPT-5.6 Sol price cut analysis, GPT-5.6 Sol direct pricing breakdown, OpenRouter alternatives