OpenAI vs Anthropic: what LLM API costs actually differ

The rate cards are nearly a wash — deliberately

At the flagship tier the rate card now favors Anthropic outright: GPT-5.4 at $2.5/15 per 1M tokens against Claude Sonnet 5.5 at $2/10 — cheaper on both lines. Compare providers that way and you’ll pick Anthropic. The workload table above has OpenAI ahead anyway. Three structural differences — none of them on either pricing page — decide the real bill.

1 · The tokenizer: Claude bills more tokens for the same text

Providers meter their own tokenizers, and they don’t tokenize alike. For typical English prose, Claude’s tokenizer produces roughly 43% more tokens than OpenAI’s o200k_base for the same input. Stack that on the sticker gap and Sonnet’s effective input price lands at ~1.14× GPT-5.4’s — not the 1.2× the rate card implies. This is the single most-missed line item in provider comparisons.

2 · The cache-write premium: Anthropic charges to warm the cache

Both providers discount cached input ~90%. The difference is on the write side: Anthropic bills first-time cache writes at 1.25× input ($2.5/1M on Sonnet); OpenAI charges no write premium at all. At low cache hit rates you pay Anthropic’s premium without earning it back — the two providers’ caching economics only converge once your hit rate is high. If your traffic is spiky, that asymmetry quietly favors OpenAI.

3 · The budget tier is where the % gap is widest

This is where the two providers diverge most, and as of the Claude 5.5 line it diverges the other way. Claude Haiku 5.5 ($0.1/0.5) undercuts GPT-5.4 mini ($0.75/4.5) by roughly 87% on input and 89% on output. The tokenizer gap applies here too and still favors OpenAI, but it is nowhere near large enough to close a gap this size — Haiku lands at roughly 0.19× mini’s effective input price. High-volume budget-tier work (simple chat, classification, summarization) is where provider choice moves the most dollars in relative terms, and at this tier the answer is no longer close.

Reading the workload table honestly

On the realistic-column defaults above (8% retries, 40% cache hits, 15% infra overhead), the result splits by tier rather than by vendor. At the flagship tier OpenAI stays ahead — ~$490/mo vs ~$538/mo (+10%) for a history-carrying chatbot at 1,000 requests/day — entirely on the tokenizer, since Sonnet 5.5 is cheaper on both sticker lines. At the budget tier Anthropic wins decisively, and by a far larger margin than OpenAI’s flagship lead. Neither is a verdict — three things legitimately move the flagship comparison:

• Quality fit on your task. If Sonnet resolves your tickets in one turn where GPT-5.4 takes two, the per-token gap is irrelevant — $/completed-task is the metric that matters.
• High, steady cache hit rates. Past the write premium’s break-even, Anthropic’s caching economics are as strong as anyone’s — a stable-prefix, high-traffic workload narrows the gap materially.
• Long context. Sonnet’s 1000K-token context at standard pricing — against GPT-5.4’s 272K — changes the architecture question for document-heavy workloads, sometimes eliminating a RAG pipeline’s worth of calls entirely.