Model comparison
GPT-6 Astra vs GPT-5.5: why sessions cost 2.1x, not 2x
GPT-6 Astra doubles GPT-5.5's input price and adds a 1.25x cache-write charge GPT-5.5 never had. What that does to agentic sessions and to Codex.
· Prices as of September 28, 2026
GPT-6 Astra
OpenAI · Released September 3, 2026
OpenAI's top GPT-6 model and its most expensive, aimed at the hardest long-running work that spans many tools, including coding.
GPT-6 Astra facts and comparisonsGPT-5.5
OpenAI · Released April 23, 2026 · Previous generation
OpenAI's April 2026 flagship for coding and professional work, now a previous generation that stays in the API.
GPT-5.5 facts and comparisons
The short answer
GPT-6 Astra costs 2x GPT-5.5 for input and 1.7x for output, but 2.5x for cache writes, because OpenAI bills writes at 1.25x input from GPT-5.6 on. That pushes a cache-heavy agentic coding session to $10.50 on Astra against $5.00 on GPT-5.5, a 2.1x gap. Astra is Codex CLI's bundled default, while GPT-5.5 exits ChatGPT and Codex sign-in on October 14, 2026.
Choose GPT-6 Astra if
- You take on the hardest end-to-end work, which OpenAI names as GPT-6 Astra's purpose.
- You use Codex CLI, whose bundled model list defaults to Astra, and you lose GPT-5.5 under ChatGPT sign-in on October 14, 2026.
- You run long tasks, where OpenAI says Astra stays coherent better than GPT-5.6 Sol and earlier models.
Choose GPT-5.5 if
- You call the API directly and want half the input price: $5 per million tokens against $10.
- Your sessions write a lot to the cache, and GPT-5.5 bills written tokens as ordinary input.
- You rely on OpenAI's description of GPT-5.5 for precise tool use on large tool surfaces and long-running agent tasks.
Side by side
Specs and prices
| Fact | GPT-6 Astra | GPT-5.5 |
|---|---|---|
| Maker | OpenAI | OpenAI |
| API model id | gpt-6-astra | gpt-5.5 |
| Released | September 3, 2026 | April 23, 2026 |
| Status | Current | Previous generation |
| Context window | 1.05M tokens | 1.05M tokens |
| Max output | 128K tokens | 128K tokens |
| Open weights | No | No |
| Input, per 1M tokens | $10 | $5 |
| Cache hit, per 1M | $1 | $0.50 |
| Cache write, per 1M | $12.50 | $5 (same as input) |
| Output, per 1M tokens | $50 | $30 |
| Runs in | Codex, OpenCode, OpenRouter, and GitHub Copilot | Codex, Cursor, OpenCode, OpenRouter, and GitHub Copilot |
Standard API rates in US dollars, as published by each maker on September 28, 2026. Batch and priority tiers, taxes, and subscription plans are not included. GPT-6 Astra: Requests over 272K input tokens cost 2x for input and cache and 1.5x for output, for the whole request. GPT-5.5: Prompts over 272K input tokens cost 2x for input and 1.5x for output, for the full session.
Cost
What the same work costs
The same token counts, priced at each model’s published rates. Cached tokens are billed at each maker’s cache prices, so the session shows what caching is worth on each model.
Real sessions differ: the two models count the same code as different numbers of tokens, and reasoning settings change how much each one writes. Your own history is the real test.
| Workload | GPT-6 Astra | GPT-5.5 |
|---|---|---|
| Agentic coding session, 100K input, 400K written to cache, 2M read from cache, 50K output | $10.50 | $5.00 |
| Large one-off review, 150K input with no cache hits, 10K output | $2.00 | $1.05 |
| Output-heavy generation, 30K input, 80K output | $4.30 | $2.55 |
| A month of sessions, 110 sessions: 5 a day, 22 working days | $1,155.00 | $550.00 |
| Where the session’s cost goes | ||
| Cache writes | $5.00 | $2.00 |
| Cache reads | $2.00 | $1.00 |
| Uncached input | $1.00 | $0.50 |
| Output | $2.50 | $1.50 |
- caching saves on the session with GPT-6 Astra (62%)
- $17.00
- caching saves on the session with GPT-5.5 (64%)
- $9.00
Why cache writes widen the gap between GPT-6 Astra and GPT-5.5
Per million tokens, GPT-6 Astra costs $10 for input and $50 for output, while GPT-5.5 costs $5 and $30. Cache hits follow input at 0.1x on both, $1 against $0.50.
Cache writes are where the generations part. From GPT-5.6 on, OpenAI charges 1.25x input to write the cache, so Astra's writes cost $12.50 per million tokens. GPT-5.5 bills written tokens as ordinary input, at $5. That is a 2.5x gap on writes, wider than the 2x on input.
In the example session, 400K tokens go into the cache: $5.00 on Astra and $2.00 on GPT-5.5, a difference of $3.00. Writes make up 48% of Astra's session and 40% of GPT-5.5's. The session gap comes out at 2.1x, $10.50 against $5.00, while the output-heavy generation, which writes nothing to the cache, shows 1.7x at $4.30 against $2.55. Here caching widens the gap instead of narrowing it.
What OpenAI built each flagship for
OpenAI launched GPT-5.5 in April 2026 as a new class of intelligence for coding and professional work. It claimed strong results with fewer reasoning tokens than earlier models at the same effort, complex coding that needs planning, tool use, codebase navigation, and verification, and precise tool use across large tool surfaces.
GPT-6 Astra arrived on September 3, 2026, as OpenAI's top GPT-6 model and its most expensive. OpenAI calls it its most capable model, built for the hardest end-to-end work, and says it stays coherent during long tasks better than GPT-5.6 Sol and earlier models. It also says Astra needed substantially fewer output tokens for stronger results in several evaluations, which it frames as a lower estimated cost per task.
Both efficiency claims concern output length, which matters because output is 24% of Astra's session cost and 30% of GPT-5.5's. If Astra writes fewer tokens on your tasks, the per-task gap shrinks below the per-token one. Only your own history can show whether it does.
What happens to GPT-5.5 in Codex on October 14
On October 14, 2026, GPT-5.5 drops out of ChatGPT and Codex sign-in, while the API keeps it. Codex CLI's bundled model list, version 0.158.0, already makes GPT-6 Astra the default and starts it at low reasoning effort. In Codex, Astra is available in the app, the CLI, and the IDE extension, though Codex cloud does not offer it.
Moving off GPT-5.5 because of that date does not have to mean Astra. The Codex docs recommend GPT-6 Sol for complex coding, and Sol is priced well below Astra.
Their published context windows and output limits match, at 1.05M of context and 128K of output on each. Long prompts cost more on both: past 272K input tokens, input costs 2x and output 1.5x, applied to the whole request on Astra and to the full session on GPT-5.5. EveryToken reads your local Codex history, prices each request at API rates, and shows what caching saved or cost per model, which makes the write premium visible.
Prompt caching
How OpenAI bills cached tokens
OpenAI
Prompt caching is on by default. From GPT-5.6 on you can also mark up to four explicit cache breakpoints, while GPT-5.5 and earlier cache automatically only.
From GPT-5.6 on, a cache write costs 1.25x the uncached input price and a cache hit costs 0.1x. GPT-5.5 and earlier add no charge for writing the cache: written tokens are billed as ordinary input, and a hit costs 0.1x on the models compared here.
On GPT-5.6 and later, a cached prefix stays reusable for at least 30 minutes after its last use, and caching starts at 1,024 input tokens.
Source: OpenAI: Prompt caching
Your own numbers
See what GPT-6 Astra and GPT-5.5 really cost you.
everyaitoken reads your Codex, OpenCode, OpenRouter, and Cursor history on your Mac and prices every request at API rates, with what caching saved or cost. $9 once.
FAQ
Questions
Is GPT-6 Astra twice the price of GPT-5.5?
For input and cache hits, yes. Output is 1.7x and cache writes 2.5x, so a cache-heavy session costs 2.1x: $10.50 against $5.00 in the example. Over 110 sessions a month that is $1,155.00 against $550.00.
Why does GPT-6 Astra charge for cache writes when GPT-5.5 doesn't?
OpenAI's caching prices changed with GPT-5.6. From GPT-5.6 on, a cache write costs 1.25x input and a hit 0.1x, while GPT-5.5 and earlier bill written tokens as ordinary input.
Does caching save more on GPT-5.5?
As a share, slightly: 64% of the uncached session on GPT-5.5 against 62% on GPT-6 Astra. In dollars Astra saves more, $17.00 against $9.00, because its input price is higher.
Is GPT-5.5 still available after October 14, 2026?
Yes, in the API. It leaves ChatGPT and Codex sign-in on that date.
Sources
- OpenAI: API pricing
- OpenAI docs: GPT-6 Astra
- OpenAI: Using the latest model
- OpenAI: API changelog
- Codex docs: Models
- Codex CLI: bundled model catalog
- OpenRouter: GPT-6 Astra
- OpenCode docs: Zen
- GitHub Docs: Supported AI models in Copilot
- OpenAI docs: GPT-5.5
- OpenAI: Using GPT-5.5
- Cursor docs: Models and pricing
- OpenRouter: GPT-5.5
- OpenAI: Prompt caching