Claude Sonnet 5.5 vs GPT-6.1 Sol: same price, what actually differs

Updated October 8, 2026. Researched and drafted with AI help; every figure is checked against the linked source.

Within two days of each other, Anthropic and OpenAI replaced their mid-tier models. Claude Sonnet 5.5 came out on September 28, 2026, and GPT-6.1 Sol followed on September 29 at OpenAI’s DevDay. Both cost exactly $2 per million input tokens and $10 per million output tokens, the same as the models they replace. So the headline price tells you nothing. This guide looks at the parts that do differ: long prompts, cache pricing history, and what each company claims about quality.

The prices side by side

All prices are per million tokens, from each provider’s official pricing page.

	Claude Sonnet 5.5	GPT-6.1 Sol
Input	$2	$2
Output	$10	$10
Cache read (cached input)	$0.10 (was $0.20 until Oct 7)	$0.10
Cache write	$2.50 (5-minute cache)	$2.50
Prompts over 272K tokens	Same price	$4 input / $0.20 cached / $15 output
Context window	1M tokens	1.05M tokens
Batch API	50% off	50% off
Released	Sep 28, 2026	Sep 29, 2026

Two rows matter:

Cache reads now cost the same. OpenAI cut cached input from $0.20 on GPT-6 Sol to $0.10, which it says is 95% below the normal input price. Sonnet 5.5 launched at $0.20, like Sonnet 5 before it, but on October 7, 2026 Anthropic halved it to $0.10, saying this cuts costs on most agentic tasks by about 20%.
Long prompts cost more only on GPT-6.1 Sol. OpenAI lists a separate long-context price of $4 input, $0.20 cached input and $15 output. Anthropic says Claude 4.6 and later models bill the full 1M-token context window at the standard rate, so “a 900k-token request is billed at the same per-token rate as a 9k-token request.”

What 1,000 requests cost

We use the same three workloads as our cost per 1,000 requests table:

Chat: 2,000 input tokens and 500 output tokens.
Coding agent step: 2,000 fresh input tokens, 18,000 tokens read from the prompt cache, and 1,000 output tokens.
Summary: 10,000 input tokens and 300 output tokens.

Workload (1,000 requests)	Claude Sonnet 5.5	GPT-6.1 Sol
Chat	$9.00	$9.00
Coding agent step	$15.80	$15.80
Summary	$23.00	$23.00
Agent step via Batch API (50% off)	$7.90	$7.90

With identical input, output and cache read prices, all three workloads cost the same on paper. Before Anthropic’s October 7 cache-read cut, the 18,000 cached tokens cost $3.60 per 1,000 steps on Sonnet 5.5 against $1.80 on GPT-6.1 Sol, which put Sonnet 5.5 at $17.60 per 1,000 agent steps. Now both cost $1.80. In practice the bills can still differ because the two models count tokens differently and may use different amounts of output for the same task.

Long prompts go the other way. Take one request with 400,000 input tokens (a large codebase or a long contract bundle) and 4,000 output tokens, without caching:

Claude Sonnet 5.5: 400,000 × $2 + 4,000 × $10 per million = $0.84
GPT-6.1 Sol at the long-context rate: 400,000 × $4 + 4,000 × $15 per million = $1.66

So for very long single prompts, Sonnet 5.5 costs about half as much.

Token counts are not equal either. The same text does not produce the same number of tokens on both models: Anthropic says Claude 4.7 and later models use a newer tokenizer that produces about 30% more tokens than earlier Claude models. Run a few of your real prompts through both APIs and compare the billed token counts before you trust any per-token comparison, including ours.

What each company claims about quality

Both companies published benchmarks, but they chose different tests and compared against different models. Neither set is independent, and Anthropic’s comparisons are against GPT-6 Sol, not GPT-6.1 Sol. Treat these as the makers’ own claims.

Anthropic on Sonnet 5.5:

Terminal-Bench 4.0 (agentic coding): 70.6%, against 10.3% for Sonnet 5 and 66.4% for Claude Opus 5.5.
GDPval-AA v2.1 (knowledge work): 1844, against 1846 for Opus 5.5 and 1487 for GPT-6 Sol.
OSWorld 2.1 partial (computer use): 80.1%, against 81.8% for Opus 5.5.
Output is “30%+ faster” than Sonnet 5, and Anthropic says it “costs up to 30% less per task than its predecessor” in its testing.

OpenAI on GPT-6.1 Sol:

DeepSWE v1.1 (software engineering): matches GPT-6 Astra “at roughly one-fifth of the cost” and beats GPT-6 Sol by 6.4 percentage points.
OSWorld 2.0 (computer use): seven percentage points above GPT-6 Sol at maximum reasoning effort.
AutomationBench (multi-step workflows): 2.2 percentage points above Opus 5.5 at medium effort.
Terminal-Bench Science 0.1: an average of $5.47 per task, against $23.21 for Opus 5.5 and $23.80 for GPT-6 Astra.
At low reasoning effort, the share of responses with a factual error falls from 11.4% (GPT-6 Sol) to 7.7%.

Notice that each company measures computer use on a different OSWorld version, and coding on a different benchmark. There is no shared number here that lets you rank the two directly.

Where you can use them

Claude Sonnet 5.5 is available in the Claude apps, through the Claude API as claude-sonnet-5-5, and on Amazon Web Services, Google Cloud and Microsoft Azure. Anthropic released the smaller Claude Haiku 5.5 on October 7, 2026.
GPT-6.1 Sol is in the API as gpt-6.1-sol and, in ChatGPT, for Plus, Pro, Business, Enterprise and Edu users in ChatGPT Work and Codex. OpenAI says it “is not yet available in Chat.” A faster “Ultrafast” version for Codex is announced but not priced yet. OpenAI’s GPT-6 Sol model page now points to GPT-6.1 Sol as the newer Sol model.

Which one to pick

Agents that re-read the same context every turn: no list-price difference any more, since both charge $0.10 for cache reads. Measure the tokens each model actually uses on your own agent runs.
Single very long prompts (over 272K tokens): Sonnet 5.5, because Anthropic doesn’t charge a long-context premium.
Plain chat, writing and summaries: the list prices are identical, so pick on quality for your own tasks and on how many tokens each model actually uses for them.
Already on Sonnet 5 or GPT-6 Sol: both upgrades keep the same input and output price, and both now charge $0.10 for cache reads, half of what Sonnet 5 and GPT-6 Sol charge. There is no price reason to stay on the old version.

To try your own token counts, cache share and batch setting, use the cost calculator, or open the Sonnet 5.5 vs GPT-6.1 Sol comparison. New models often get “it got worse” reports in their first weeks, so check the live community votes on each model page and the dumb meter next to the official incident list. You can also embed a live rating badge for either model from the badges page.