OpenAI vs Anthropic
API pricing and model line-up compared · Prices checked 12 August 2026
OpenAI has the lower entry price: GPT-4.1 nano at $0.190 per million tokens blended, against Claude Haiku 4.5 at $1.76 on Anthropic, roughly 9.3× cheaper to start.
At the top of the range Anthropic goes higher: Claude Opus 4.5 at $33.00 per million tokens against o1 at $28.50. OpenAI offers more choice in our catalogue: 17 text models against 13. OpenAI offers the larger maximum context window, 1,050,000 tokens against 1,000,000.
Catalogue side by side
| Metric | OpenAI | Anthropic |
|---|---|---|
| Models tracked | 17 | 13 |
| Cheapest model | GPT-4.1 nano · $0.190/M | Claude Haiku 4.5 · $1.76/M |
| Most expensive model | o1 · $28.50/M | Claude Opus 4.5 · $33.00/M |
| Top-tier input / output | $5.00 / $30.00 | $5.00 / $25.00 |
| Largest context window | 1,050,000 | 1,000,000 |
| Models with public benchmarks | 14 of 17 | 10 of 13 |
| Founded | 2015 | 2021 |
| Headquarters | San Francisco, USA | San Francisco, USA |
Counts cover text-generation models in our curated catalogue, excluding retired SKUs, embeddings and non-text products. Blended rates weight input at 70 per cent and output at 30 per cent; how that is calculated.
About each provider
OpenAI
Maker of ChatGPT, GPT and the o-series reasoning models. Broadest API surface in the industry: chat, vision, audio, embeddings, image generation.
Top of the line: GPT-5.6 Sol at $5.00 input / $30.00 output per million tokens, 1,050,000 context.
Anthropic
AI safety company building Claude — the assistant family known for long context, careful reasoning and best-in-class coding via Claude Code.
Top of the line: Claude Opus 5 at $5.00 input / $25.00 output per million tokens, 1,000,000 context.
Frequently asked questions
Is OpenAI or Anthropic cheaper?
It depends which tier you need. OpenAI has the lower entry price, but both run a range from budget to frontier models, so the fair comparison is tier against tier rather than vendor against vendor. Price the specific models you would actually call; the table above shows each side’s floor and ceiling.
Should I use OpenAI or Anthropic?
Neither wins outright. Shortlist by the capability tier your workload needs, then compare the specific models on price, context window and benchmark evidence. Multi-vendor setups are common precisely because the cheapest adequate model for one task is rarely the same vendor as for another.
Where do these OpenAI and Anthropic prices come from?
Each provider’s published rates, as tracked in our pricing index. Blended figures weight input at 70 per cent and output at 30 per cent and are calculated comparison metrics rather than published prices. Counts cover text-generation models only, excluding retired SKUs, embeddings and non-text products.
Go deeper
Other provider matchups
Figures are recomputed on every deploy from the curated catalogue. Prices checked 12 August 2026.