Model comparison
Claude Sonnet 5 vs GPT-5.5: API costs and the Codex exit
GPT-5.5 costs 2.5x Claude Sonnet 5 for input and 3x for output, and it leaves ChatGPT and Codex sign-in on October 14, 2026. What that means for coding.
· Prices as of September 28, 2026
Claude Sonnet 5
Anthropic · Released June 30, 2026
Anthropic's balance of speed and intelligence, and a drop-in upgrade for Claude Sonnet 4.6. Anthropic pitches it as close to Claude Opus 4.8 at a lower price.
Claude Sonnet 5 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
Claude Sonnet 5 costs less on every workload compared here: the example agentic coding session is $2.40 against $5.00 on GPT-5.5, and output-heavy generation is $0.86 against $2.55. GPT-5.5 is OpenAI's April 2026 flagship, now a previous generation that leaves ChatGPT and Codex sign-in on October 14, 2026, while staying in the API. It mainly suits API users whose prompts are already tuned for it.
Choose Claude Sonnet 5 if
- You want lower rates: $2 input and $10 output per million, against $5 and $30 on GPT-5.5.
- Your work is output-heavy, where GPT-5.5 costs 3x as much.
- Claude Code is your main tool, and its sonnet alias selects Sonnet 5 on the Anthropic API.
- You want explicit cache breakpoints, which Claude supports and GPT-5.5 does not.
Choose GPT-5.5 if
- You call GPT-5.5 through the API and your prompts and tooling are tuned to it.
- Your agents work across large tool surfaces, where OpenAI names precise tool use and long-running agent tasks as GPT-5.5 strengths.
Side by side
Specs and prices
| Fact | Claude Sonnet 5 | GPT-5.5 |
|---|---|---|
| Maker | Anthropic | OpenAI |
| API model id | claude-sonnet-5 | gpt-5.5 |
| Released | June 30, 2026 | April 23, 2026 |
| Status | Current | Previous generation |
| Context window | 1M tokens | 1.05M tokens |
| Max output | 128K tokens | 128K tokens |
| Open weights | No | No |
| Input, per 1M tokens | $2 | $5 |
| Cache hit, per 1M | $0.20 | $0.50 |
| Cache write, per 1M | $2.50 (5-minute), $4 (1-hour) | $5 (same as input) |
| Output, per 1M tokens | $10 | $30 |
| Runs in | Claude Code, Cursor, OpenCode, OpenRouter, and GitHub Copilot | Codex, Cursor, OpenCode, OpenRouter, and GitHub Copilot |
Standard API rates in US dollars, as published by each maker (Claude Sonnet 5: September 26, 2026; GPT-5.5: September 28, 2026). Batch and priority tiers, taxes, and subscription plans are not included. Claude Sonnet 5: The full 1M context window is billed at standard rates. The launch price became the standard price on August 10, 2026, and a planned increase was cancelled. 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 | Claude Sonnet 5 | GPT-5.5 |
|---|---|---|
| Agentic coding session, 100K input, 400K written to cache, 2M read from cache, 50K output | $2.40 | $5.00 |
| Large one-off review, 150K input with no cache hits, 10K output | $0.40 | $1.05 |
| Output-heavy generation, 30K input, 80K output | $0.86 | $2.55 |
| A month of sessions, 110 sessions: 5 a day, 22 working days | $264.00 | $550.00 |
| Where the session’s cost goes | ||
| Cache writes | $1.30 | $2.00 |
| Cache reads | $0.40 | $1.00 |
| Uncached input | $0.20 | $0.50 |
| Output | $0.50 | $1.50 |
- caching saves on the session with Claude Sonnet 5 (56%)
- $3.10
- caching saves on the session with GPT-5.5 (64%)
- $9.00
Why the gap grows with output
GPT-5.5 charges $5 per million input tokens and $30 per million output tokens. Claude Sonnet 5 charges $2 and $10. Input is 2.5x apart and output 3x apart, so the more a job writes, the wider the gap gets.
The three workloads show it. The large one-off review, mostly input, costs $1.05 against $0.40, a 2.6x gap. The output-heavy generation costs $2.55 against $0.86, 3x. The agentic session, dominated by cached context, costs $5.00 against $2.40, the narrowest gap at 2.1x.
The session gap is narrow because of how each maker bills cache writes. GPT-5.5 adds no write charge, so its written tokens cost $5 per million, which is only 1.25x Sonnet 5's 1-hour write price and 2x its 5-minute price. Writes cost $2.00 on GPT-5.5 against $1.30 on Sonnet 5. The largest single difference in the session is output: $1.50 against $0.50. Over 110 sessions a month the totals are $550.00 and $264.00.
What changes for GPT-5.5 on October 14, 2026
On October 14, 2026, GPT-5.5 is removed from ChatGPT and Codex sign-in. It stays in the OpenAI API, and it is currently offered in Codex, Cursor, OpenRouter, OpenCode, and GitHub Copilot. If you use it in Codex with a ChatGPT account, you will need another model after that date.
OpenAI launched GPT-5.5 in April 2026 as "a new class of intelligence for coding and professional work." It says the model reaches strong results with fewer reasoning tokens than earlier models at the same effort, and handles complex coding that needs planning, tool use, codebase navigation, verification, and multi-step execution.
Anthropic released Sonnet 5 at the end of June 2026 and calls it "built to be the most agentic Sonnet model yet," with performance close to Claude Opus 4.8. Its launch price became the standard price on August 10, 2026.
Caching and long prompts work differently
GPT-5.5 predates OpenAI's explicit cache breakpoints, which arrived with GPT-5.6. Its caching is automatic only: there are no breakpoints to mark, no write charge, and a cache hit costs 0.1x input, $0.50 per million. Claude lets a tool place breakpoints, and Claude Code manages them itself. Caching saves $9.00 on the GPT-5.5 session, 64% of the uncached cost, against $3.10 and 56% on Sonnet 5.
Long prompts cost more on GPT-5.5. Once a prompt goes over 272K input tokens, input costs 2x and output 1.5x for the full session, not only that request. The example session stays well below that. Context windows are 1.05M on GPT-5.5 and 1M on Sonnet 5, and both write up to 128K tokens.
Neither tokenizer counts text the same way, and reasoning effort changes how much each model writes. Treat the figures here as API-equivalent estimates at published rates. EveryToken can show what your own Claude Code and Codex requests cost on each model.
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 Sonnet 5 and GPT-5.5 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-5.5 more expensive than Claude Sonnet 5?
Yes, on every rate: 2.5x for input and cache hits, 3x for output, and 2x for a cache write compared with Sonnet 5's 5-minute write. The example session costs $5.00 against $2.40, a $2.60 difference.
Can I still use GPT-5.5 after October 14, 2026?
It leaves ChatGPT and Codex sign-in on that date and stays available in the OpenAI API. Plan a switch if you use it through a ChatGPT account in Codex.
How does GPT-5.5 bill prompts above 272K tokens?
Input is billed at 2x and output at 1.5x for the full session once a prompt passes 272K input tokens. Claude Sonnet 5 has a 1M context window, and GPT-5.5 has 1.05M.
Is there a cache-write premium on GPT-5.5?
No. GPT-5.5 and earlier OpenAI models bill written tokens as ordinary input and cache automatically. From GPT-5.6 on, OpenAI charges 1.25x input for a write and adds explicit breakpoints.
Sources
- Anthropic: Pricing
- Anthropic docs: Claude Sonnet 5
- Anthropic: Introducing Claude Sonnet 5
- Claude Code docs: Model configuration
- Cursor docs: Claude Sonnet 5
- OpenRouter: Claude Sonnet 5
- OpenCode docs: Zen
- GitHub Docs: Supported AI models in Copilot
- OpenAI: API pricing
- OpenAI docs: GPT-5.5
- OpenAI: Using GPT-5.5
- OpenAI: API changelog
- Codex docs: Models
- Cursor docs: Models and pricing
- OpenRouter: GPT-5.5
- Anthropic: Prompt caching
- OpenAI: Prompt caching