# GPT-5.6 Sol > GPT-5.6 Sol per million tokens, with a long-context tier, cache read and a 5-minute cache write. Agent-loop cost at 0%, 50% and 90% hit, next to GPT-5.6 Luna. - URL: https://bytescope.dev/en/llm-pricing/gpt-5.6-sol/ - Locale: en · Also: [ja](https://bytescope.dev/ja/llm-pricing/gpt-5.6-sol.md) · [zh-Hant](https://bytescope.dev/zh-Hant/llm-pricing/gpt-5.6-sol.md) - Runs: The rate card is read in this tab. Nothing is uploaded and no model API is called. - Requires: Any modern browser (Chrome, Edge, Firefox, Safari), desktop and mobile. - Updated: 2026-09-08 ## What it does ### GPT-5.6 Sol Sol is the id this table also accepts under the plain gpt-5.6 name. It sits between Astra and Luna, and it has the same shape as both: a long-context tier, a cache read, a five-minute cache write, no one-hour write, and a batch factor the calculator does not apply. The page's job is the spread between Sol and Luna on the same loop, which is the decision when the task fits the cheaper card and you want the delta in money rather than a vibe. Astra remains the flagship comparison if the question is whether the top card is worth it. Luna is the one this page subtracts, because that is the switch people actually hesitate over. ### How this card is billed The numbers on this page are not typed in. A checked rate card lives in one table, and every figure you see is read from that table or returned by the cost function the rest of the site calls. The scenario is `agent-loop-50`: fifty turns, one cacheable context that comes back every turn, a little new input that is never cached, and output on every turn. The three totals further down set the cache hit rate at zero, one half, and nine tenths. Hit rate is the share of that repeated context that is read from cache instead of written again. It is not a discount applied after the fact, and it is not the batch factor. Batch, when the card has one, is listed on the card and is left out of these totals on purpose, because a batch job is a different purchase from the interactive loop. When a card publishes no price for a cache line the scenario still has to account for, those tokens are billed as fresh input and the result is marked so. That is a property of the card, not a missing feature of the page. A long-context tier, when the card has one, applies to a single request whose prompt is over the threshold. The fifty short turns of this loop stay on the base tier, which is why a model with a long-context tier still shows one set of loop totals. The worked example below names the day the card was confirmed against the vendor page linked next to it. If you need a different loop, the parent page has the other three scenarios and a slider. ### Worked example Rates on this page were checked against the vendor pricing page on 2026-09-22. 0% cache hit: $11.10 50% cache hit: $6.50 90% cache hit: $2.82 Versus GPT-5.6 Luna: $6.17 more than GPT-5.6 Luna. ### Long context A prompt over 272,000 tokens uses the long-context rates. One request with 272,001 input tokens and 1,000 output tokens, and no cache tokens, costs $2.21. ## Limits Does not call a model API, does not read agent logs, and does not apply the batch factor to the scenario totals. Rates are the checked card, not a live fetch. ## FAQ ### What does the date on the GPT-5.6 Sol page mean? It is the day this card was checked against the vendor pricing page. It is not the model's release day, and it is not a promise that the vendor has not changed a number since. When the card changes, the date changes with it. ### Which cache lines does GPT-5.6 Sol publish? This card publishes cache read and a five-minute cache write, and it does not publish a one-hour cache write. A missing one-hour line is not a free hour of cache. It also publishes a long-context tier above a prompt threshold, and a batch factor. The loop totals stay on the base tier. ### Why is GPT-5.6 Luna the comparison? Luna keeps Sol's shape and lowers the schedule. That is the switch to price. Astra is a different question and has its own page.