Model comparison
Claude Sonnet 5 vs GPT-5.6 Sol: price, promo, and successor
GPT-5.6 Sol charges twice Claude Sonnet 5's rates on promotional pricing that runs through at least November 21, 2026. A cached session: $4.20 vs $2.40.
· 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-5.6 Sol
OpenAI · Released July 9, 2026 · Previous generation
The flagship of the July 2026 GPT-5.6 family, now succeeded by GPT-6 Sol and on promotional pricing.
GPT-5.6 Sol facts and comparisons
The short answer
Claude Sonnet 5 costs half as much per token as GPT-5.6 Sol, and the example agentic coding session comes to $2.40 against $4.20, a 1.8x gap. GPT-5.6 Sol is a previous-generation model on promotional rates available at least through November 21, 2026, and outside Codex cloud the Codex docs suggest GPT-6 Sol in its place. Stay on GPT-5.6 Sol mainly if Codex cloud or code that calls the gpt-5.6 API id depends on it.
Choose Claude Sonnet 5 if
- You want half the per-token price: $2 input and $10 output per million against $4 and $20.
- You want rates without an end date attached: Sonnet 5's launch price became its standard price on August 10, 2026.
- You code in Claude Code and want the model its sonnet alias selects on the Anthropic API.
- You would rather adopt a current model than one its maker has already succeeded.
Choose GPT-5.6 Sol if
- Your team chats with Codex cloud on a ChatGPT plan, and those chats use GPT-5.6 Sol.
- Your code calls the gpt-5.6 API id, which points to GPT-5.6 Sol, and you are not ready to migrate.
- You build frontends, where OpenAI credits GPT-5.6 with better layout, visual hierarchy, and design judgment.
- You want programmatic tool calling, in which the model writes JavaScript that calls tools and processes their output.
Side by side
Specs and prices
| Fact | Claude Sonnet 5 | GPT-5.6 Sol |
|---|---|---|
| Maker | Anthropic | OpenAI |
| API model id | claude-sonnet-5 | gpt-5.6-sol |
| Released | June 30, 2026 | July 9, 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 | $4 |
| Cache hit, per 1M | $0.20 | $0.40 |
| Cache write, per 1M | $2.50 (5-minute), $4 (1-hour) | $5 |
| Output, per 1M tokens | $10 | $20 |
| 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-5.6 Sol: 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-5.6 Sol: Requests over 272K input tokens cost 2x for input and cache and 1.5x for output, for the whole request. These are promotional rates, available at least through November 21, 2026.
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-5.6 Sol |
|---|---|---|
| Agentic coding session, 100K input, 400K written to cache, 2M read from cache, 50K output | $2.40 | $4.20 |
| Large one-off review, 150K input with no cache hits, 10K output | $0.40 | $0.80 |
| Output-heavy generation, 30K input, 80K output | $0.86 | $1.72 |
| A month of sessions, 110 sessions: 5 a day, 22 working days | $264.00 | $462.00 |
| Where the session’s cost goes | ||
| Cache writes | $1.30 | $2.00 |
| Cache reads | $0.40 | $0.80 |
| Uncached input | $0.20 | $0.40 |
| Output | $0.50 | $1.00 |
- caching saves on the session with Claude Sonnet 5 (56%)
- $3.10
- caching saves on the session with GPT-5.6 Sol (62%)
- $6.80
GPT-5.6 Sol's promotional rates and what they cost
GPT-5.6 Sol currently lists $4 per million input tokens, $0.40 per million cache hits, $5 per million cache writes, and $20 per million output tokens. OpenAI marks these as promotional rates, available at least through November 21, 2026. The price data behind this post does not say what replaces them, so check OpenAI's pricing page before planning past that date.
At today's rates every token type costs twice what Claude Sonnet 5 charges. The large one-off review costs $0.80 against $0.40, and the output-heavy generation $1.72 against $0.86. Sonnet 5's pricing moved the other way in 2026: its launch price became the standard price on August 10, 2026, and a planned increase was cancelled.
Why the session gap is 1.8x rather than 2x
Three of the session's four cost lines are exactly double on GPT-5.6 Sol: cache reads at $0.80 against $0.40, fresh input at $0.40 against $0.20, and output at $1.00 against $0.50. Cache writes break the pattern. The example session writes 400K tokens, and on Claude half of them go to the 1-hour cache at $4 per million, 2x the input price. GPT-5.6 Sol charges its single $5 write rate for everything, so writes come to $2.00 against $1.30, a $0.70 gap instead of a doubling.
The totals are $4.20 and $2.40. At 110 sessions a month that is $462.00 against $264.00, a $198.00 difference in API-equivalent cost. Caching saves $6.80 on GPT-5.6 Sol, 62% of what the same tokens would cost uncached, against $3.10 and 56% on Sonnet 5.
OpenAI's caching from GPT-5.6 on works differently from Anthropic's. It is on by default, you can mark up to four explicit breakpoints, caching starts at 1,024 input tokens, and a cached prefix stays reusable for at least 30 minutes after its last use. By Anthropic's rule of thumb, one cache read repays a 5-minute write and two repay a 1-hour write.
Should you move to GPT-6 Sol instead?
OpenAI already treats GPT-5.6 Sol as the previous flagship. Codex cloud chats on ChatGPT plans still use it, but elsewhere Codex suggests moving to GPT-6 Sol, which lists the same per-token rates as Claude Sonnet 5. For anyone comparing on price alone, that makes the 2x gap in this post a gap between Sonnet 5 and an older OpenAI model.
Reasons to stay on GPT-5.6 Sol are about fit rather than cost. OpenAI pitches it on token efficiency, reaching flagship-level performance with fewer output tokens, and on frontend work and programmatic tool calling. It runs in Codex, Cursor, OpenRouter, OpenCode, and GitHub Copilot. Anthropic pitches Sonnet 5 as its most agentic Sonnet, with performance close to Claude Opus 4.8, and it runs in Claude Code, Cursor, OpenRouter, OpenCode, and GitHub Copilot.
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-5.6 Sol 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
How long do GPT-5.6 Sol's promotional prices last?
They are available at least through November 21, 2026, according to OpenAI. The rate that follows is not part of the pricing this post uses, so check OpenAI's pricing page as that date approaches.
Is GPT-5.6 Sol still available in Codex?
Yes. Codex cloud chats on ChatGPT plans use it, and the gpt-5.6 API id points to it. In other Codex surfaces the docs suggest switching to GPT-6 Sol.
Do Claude Sonnet 5 and GPT-5.6 Sol have the same context window?
Nearly. Sonnet 5 has 1M tokens and GPT-5.6 Sol 1.05M, and both write up to 128K tokens of output. On GPT-5.6 Sol, requests over 272K input tokens cost 2x for input and cache and 1.5x for output.
How do I see which model costs my team more?
EveryToken reads Claude Code, Codex, and Cursor history on your Mac and prices each request at API rates, with cache savings per model. Its totals are estimates at published rates, not what a ChatGPT or Claude plan charges.
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-5.6 Sol
- OpenAI: Using GPT-5.6
- OpenAI: API changelog
- Codex docs: Models
- Codex docs: Pricing
- Cursor docs: Models and pricing
- OpenRouter: GPT-5.6 Sol
- Anthropic: Prompt caching
- OpenAI: Prompt caching