Cost by workload size
Price comparison
| Feature |  GPT-4o |  Claude Sonnet 4.6 |
|---|---|---|
| Provider | OpenAI | Anthropic |
| Input / M tokens | $2.50 | $3 |
| Output / M tokens | $10 | $15 |
| Cached input read | $1.25 | $0.30 |
| Context window | 128K | 1M |
| Batch discount | −50% | — |
| Vision | Yes | Yes |
| Audio | Yes | — |
| Reasoning tokens | — | — |
What your workload actually costs
Prices per token hide the real picture. Here are three common workloads computed with AITokenCalculator's own engine (medium reasoning effort, no batch, no cache):
| Workload | GPT-4o | Claude Sonnet 4.6 | Cheaper |
|---|---|---|---|
| 1,000 in / 300 out | $0.0055 | $0.0075 | GPT-4o |
| 10,000 in / 2,000 out | $0.045 | $0.06 | GPT-4o |
| 100,000 in / 20,000 out | $0.45 | $0.6 | GPT-4o |
| Monthly @ 1,000 req/day (standard) | $1,369.69 | $1,826.25 | GPT-4o |
Analysis
On input price, GPT-4o wins at $2.50/M — roughly 1.2× cheaper than its rival. Output tells a sharper story: GPT-4o charges $10/M , about 1.5× less than Claude Sonnet 4.6 — and output is where chatty and agentic workloads bleed money.
Context capacity differs too: Claude Sonnet 4.6 fits 1M tokens (~750,000 words) versus 128K. If you feed whole documents or codebases, that gap decides feasibility before price even matters.
The default flagship showdown. OpenAI wins on price and ecosystem breadth; Claude typically wins on long-document reasoning, coding quality and safer refusals. Teams doing RAG over big corpora lean Claude; product teams needing multimodal + wide tooling lean GPT-4o.
Numbers shift as providers reprice — check the full AI model pricing table, then run your own prompt through the calculator preloaded with GPT-4o or with Claude Sonnet 4.6.