Model comparison
Claude Opus 5.5 vs GPT-6 Astra: two tool defaults, priced
Claude Code defaults to Claude Opus 5.5 and Codex CLI to GPT-6 Astra. Astra costs 2.5x more per token and 2.4x more on a cached coding session. Here is why.
· Prices as of September 28, 2026
Claude Opus 5.5
Anthropic · Released September 22, 2026
Anthropic's recommended starting model for most work, built for long-running agentic coding. Anthropic says it matches Claude Fable 5.1 on most work at a much lower price.
Claude Opus 5.5 facts and comparisonsGPT-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 comparisons
The short answer
Claude Opus 5.5 costs less: each GPT-6 Astra list price is 2.5x the Opus 5.5 rate, and the example agentic coding session costs $4.40 against $10.50, a 2.4x gap. Each is its tool's default, Opus 5.5 in Claude Code and GPT-6 Astra in Codex CLI's bundled model list, so for many developers the choice follows the tool. Anthropic says Opus 5.5 performs at the level of Claude Fable 5.1 on most work, and OpenAI pitches Astra for the hardest end-to-end work.
Choose Claude Opus 5.5 if
- You code in Claude Code, which already defaults to Opus 5.5 on Pro, Max, Team, and Enterprise plans and with an API key.
- You want the lower price: $4 input and $20 output per million tokens, 60% less than Astra.
- You want a faster option on the Claude API: fast mode, a research preview, costs $8 input and $40 output, still under Astra's standard rates.
- Your jobs are long and sprawling, like codebase-wide migrations and audits, which Anthropic names as a particular strength.
Choose GPT-6 Astra if
- You work in Codex and want its CLI default, which also runs in the Codex app and IDE extension.
- OpenAI's pitch fits your tasks: the hardest end-to-end work across many tools, where it says Astra stays coherent over long tasks better than GPT-5.6 Sol and earlier models.
- You want to test OpenAI's claim that Astra reaches stronger results with substantially fewer output tokens, which it says lowers the estimated cost per task.
Side by side
Specs and prices
| Fact | Claude Opus 5.5 | GPT-6 Astra |
|---|---|---|
| Maker | Anthropic | OpenAI |
| API model id | claude-opus-5-5 | gpt-6-astra |
| Released | September 22, 2026 | September 3, 2026 |
| Status | Current | Current |
| Context window | 1M tokens | 1.05M tokens |
| Max output | 128K tokens | 128K tokens |
| Open weights | No | No |
| Input, per 1M tokens | $4 | $10 |
| Cache hit, per 1M | $0.20 | $1 |
| Cache write, per 1M | $5 (5-minute), $8 (1-hour) | $12.50 |
| Output, per 1M tokens | $20 | $50 |
| Runs in | Claude Code, Cursor, OpenCode, OpenRouter, and GitHub Copilot | Codex, OpenCode, OpenRouter, and GitHub Copilot |
Standard API rates in US dollars, as published by each maker (Claude Opus 5.5: September 26, 2026; GPT-6 Astra: September 28, 2026). Batch and priority tiers, taxes, and subscription plans are not included. Claude Opus 5.5: The full 1M context window is billed at standard rates. Fast mode, a research preview on the Claude API, costs $8 input and $40 output per million tokens. GPT-6 Astra: Requests over 272K input tokens cost 2x for input and cache and 1.5x for output, for the whole request.
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 | Claude Opus 5.5 | GPT-6 Astra |
|---|---|---|
| Agentic coding session, 100K input, 400K written to cache, 2M read from cache, 50K output | $4.40 | $10.50 |
| Large one-off review, 150K input with no cache hits, 10K output | $0.80 | $2.00 |
| Output-heavy generation, 30K input, 80K output | $1.72 | $4.30 |
| A month of sessions, 110 sessions: 5 a day, 22 working days | $484.00 | $1,155.00 |
| Where the session’s cost goes | ||
| Cache writes | $2.60 | $5.00 |
| Cache reads | $0.40 | $2.00 |
| Uncached input | $0.40 | $1.00 |
| Output | $1.00 | $2.50 |
- caching saves on the session with Claude Opus 5.5 (60%)
- $6.60
- caching saves on the session with GPT-6 Astra (62%)
- $17.00
Why the session gap is 2.4x, not 2.5x
Most of the price sheet moves in lockstep. GPT-6 Astra charges $10 input, $50 output, and $12.50 per million for a cache write, each 2.5x the $4, $20, and $5 of Claude Opus 5.5. The uncached review, $0.80 against $2.00, and the output-heavy generation, $1.72 against $4.30, show the same 2.5x.
Two caching rules break the pattern. Anthropic prices an Opus 5.5 cache hit at 0.05x input, $0.20 per million, while OpenAI charges 0.1x on Astra, $1, a 5x difference. Anthropic also offers a 1-hour cache write at 2x input, $8 on Opus 5.5, where OpenAI has a single write price at 1.25x.
In the session those rules pull in opposite directions. Opus 5.5 spends $0.40 on cache reads against Astra's $2.00, but its writes are relatively dearer at $2.60 against $5.00, because half of them use the 1-hour rate. The net result is $4.40 against $10.50, and $484.00 against $1,155.00 over 110 sessions a month, a $671.00 difference.
Does Opus 5.5 fast mode cost more than GPT-6 Astra?
No. Fast mode, a research preview on the Claude API, prices Opus 5.5 at $8 input and $40 output per million tokens. That is double Opus 5.5's standard rates and still below Astra's $10 and $50. Separately, Anthropic says Opus 5.5 produces output more than 30% faster than Claude Opus 5.
The cost table uses standard rates for both models. If you plan to run most of your work in fast mode, its costs sit much closer to Astra's than the table suggests.
Long prompts and the 272K threshold
OpenAI bills any Astra request above 272K input tokens at 2x for input and cache and 1.5x for output, for the whole request, and Astra accepts at most 922K input tokens of its 1.05M window. Opus 5.5 has no such tier: its full 1M context window bills at standard rates. Both write up to 128K tokens per response.
The example session never reaches either limit, since each of its requests stays under 200K input tokens. Agents that pack most of a large repository into one prompt will meet Astra's surcharge, which widens the gap well beyond the table's figures.
Effort settings and the output tokens the tables hold fixed
Opus 5.5 runs adaptive thinking at medium effort by default, and Anthropic notes that changing effort keeps the prompt cache. Codex starts Astra at low reasoning effort. OpenAI says Astra reached stronger results with substantially fewer output tokens in several evaluations, for a lower estimated cost per task, and the cost table cannot capture that because it prices the same 50K output tokens on both.
Tokenizers also differ between the makers, and Opus 5.5 uses Anthropic's newer tokenizer, which counts about 30% more tokens than earlier Claude models for the same text. The reliable comparison is on your own tasks: EveryToken reads local Claude Code and Codex history on a Mac and prices each request at API rates, so the two defaults can be compared on real sessions.
Prompt caching
How each maker bills cached tokens
Anthropic
Claude caches a prompt prefix up to a breakpoint. One top-level cache_control field places the breakpoint automatically and moves it as the conversation grows, or you can mark up to 4 blocks yourself. Claude Code manages caching for you.
A 5-minute cache write costs 1.25x the input price and a 1-hour write costs 2x. A cache hit costs 0.1x input on most models, 0.05x on Claude Opus 5.5, and 0.025x on Claude Fable 5.1 and Claude Mythos 5.1. Every hit restarts the cache lifetime at no charge.
Anthropic's rule of thumb: a 5-minute write pays for itself after one cache read, and a 1-hour write after two.
In Claude Code, the main conversation uses the 1-hour cache on a Claude subscription and the 5-minute cache with an API key. Each model has its own cache, so switching models starts over.
Source: Anthropic: Prompt caching
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 Claude Opus 5.5 and GPT-6 Astra really cost you.
everyaitoken reads your Claude Code, Cursor, OpenCode, OpenRouter, and Codex 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 more expensive than Claude Opus 5.5?
Yes. Every list price is 2.5x higher on Astra, and cache hits are 5x higher. The example agentic session costs $10.50 on Astra and $4.40 on Opus 5.5, estimates at published API rates rather than subscription prices.
Which model does Codex CLI use by default?
GPT-6 Astra, in Codex CLI's bundled model list for version 0.158.0, starting at low reasoning effort. It is also in the Codex app and IDE extension, but not in Codex cloud. Claude Code's default is Claude Opus 5.5.
Does Anthropic say Opus 5.5 matches Claude Fable 5.1?
Anthropic says Opus 5.5 "performs at the level of Claude Fable 5.1 on most work and costs 40% less to run than Opus 5." It calls Opus 5.5 its recommended starting model for most work. Those are Anthropic's claims, not results measured for this page.
Can I use both models outside their makers' own tools?
Yes. Opus 5.5 is available in Cursor, OpenCode, OpenRouter, and GitHub Copilot, and GPT-6 Astra in OpenCode, OpenRouter, and GitHub Copilot. Each model keeps its own cache, so switching between them mid-task starts caching over.
Sources
- Anthropic: Pricing
- Anthropic docs: Claude Opus 5.5
- Anthropic: Claude Opus 5.5
- Anthropic docs: Fast mode
- Claude Code docs: Model configuration
- Cursor docs: Claude Opus 5.5
- OpenRouter: Claude Opus 5.5
- OpenCode docs: Zen
- GitHub Docs: Supported AI models in Copilot
- 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
- Anthropic: Prompt caching
- OpenAI: Prompt caching