GPT-5 vs Claude Sonnet 5: API pricing compared

Head-to-head on price — the default production workhorses. On a mixed 10K-input / 2K-output request, GPT-5 is about 1.2× cheaper. The full picture below.

List prices side by side

GPT-5Claude Sonnet 5
ProviderOpenAIAnthropic
Input $/MTok$1.25$2
Cached input $/MTok$0.125 (90% off)$0.2 (90% off)
Output $/MTok$10$10
Context window391K tokens977K tokens
Max output125K tokens125K tokens

Claude Sonnet 5: Launched as introductory pricing; Anthropic has since made $2 in / $10 out the standard rate (the planned rise to $3 / $15 on 1 Sep 2026 was cancelled).

What each costs in practice

WorkloadInputOutputGPT-5Claude Sonnet 5
Short chat message500250$0.0031$0.0035
10-page document + summary7,000500$0.014$0.019
Agent / coding session50,00010,000$0.163$0.200
1M tokens in + 1M out1,000,0001,000,000$11.25$12.00

Price your exact token mix on each model's own page: the GPT-5 cost calculator and the Claude Sonnet 5 cost calculator both take custom input/output counts and a requests-per-day figure.

About GPT-5

GPT-5 is the model that unified OpenAI's two earlier lines — the GPT series and the o-series reasoning models — into one system with adjustable reasoning effort. OpenAI introduced it as its flagship for coding, agentic tasks and professional writing, and made it the default model behind ChatGPT on release.

It reads text and images, writes text, and scales its reasoning from minimal for quick lookups to high for long deliberate problem-solving, so one model covers both ends. The mini and nano variants share its interface at smaller sizes, making the family a common default for production API work.

About Claude Sonnet 5

Claude Sonnet 5 is Anthropic's middle tier, described in its documentation as the best combination of speed and intelligence in the lineup. It inherits what were flagship traits a generation earlier: the million-token context window, the large output ceiling, text-and-image input and adaptive thinking, delivered at noticeably lower latency than the Opus class.

The Sonnet line has historically been where Anthropic's production API traffic concentrates — capable enough for agents and coding assistants, fast enough for interactive products — and this generation keeps that role, with extended batch output available for the workloads that need very long generations.

Frequently asked questions

Which is cheaper, GPT-5 or Claude Sonnet 5?

On a mixed workload (10K input / 2K output tokens per request), GPT-5 costs $0.033 per request versus $0.040 for Claude Sonnet 5 — about 1.2× less. List rates: GPT-5 is $1.25/M input and $10/M output; Claude Sonnet 5 is $2/M input and $10/M output.

What does a typical chat message cost on GPT-5 vs Claude Sonnet 5?

A short chat message (about 500 input and 250 output tokens) costs roughly $0.0031 with GPT-5 and $0.0035 with Claude Sonnet 5. At 1,000 messages a day that is about $93.75 versus $105.00 per month.

Do GPT-5 and Claude Sonnet 5 have the same context window?

No. GPT-5 accepts up to 391K tokens of input; Claude Sonnet 5 accepts up to 977K. For workloads built around very large prompts, that difference can matter as much as the per-token rates.

Related

Compare all models at once ·GPT-5 pricing in detail ·Claude Sonnet 5 pricing in detail ·Count tokens in your prompt

Last updated 2026-09-03. Prices verified against each provider's official pricing page; see the methodology.