Prices at each model size
Per million tokens, standard tier, short prompts. Sizes are a rough match by price and how each provider positions the model, not a claim that two models do equally well.
| Size | OpenAI | Anthropic |
|---|---|---|
| Largest | GPT-5.5: $5 in, $30 out | Claude Opus 5.5: $4 in, $20 out |
| Large | GPT-6.1 Sol: $2 in, $10 out | Claude Sonnet 5.5: $2 in, $10 out |
| Mid | GPT-6.1 Terra: $1 in, $5 out | Claude Haiku 4.5: $1 in, $5 out |
| Small | GPT-6 Luna: $0.10 in, $0.50 out | Claude Haiku 5.5: $0.10 in, $0.50 out |
The same workload on both
The example is 20 million input tokens and 4 million output tokens a month, prompts around 4,000 tokens, at list price.
| Model | Input | Output | Per month |
|---|---|---|---|
| GPT-5.5 | $100 | $120 | $220 |
| Claude Opus 5.5 | $80.00 | $80.00 | $160 |
| GPT-6.1 Sol | $40.00 | $40.00 | $80.00 |
| Claude Sonnet 5.5 | $40.00 | $40.00 | $80.00 |
| GPT-6.1 Terra | $20.00 | $20.00 | $40.00 |
| GPT-5 mini | $5.00 | $8.00 | $13.00 |
| Claude Haiku 5.5 | $2.00 | $2.00 | $4.00 |
| GPT-6 Luna | $2.00 | $2.00 | $4.00 |
Caching and batch rules differ
OpenAI caches repeated prompt starts on its own and charges nothing to write the cache. Anthropic caches what you mark, charges 1.25 times the input price to write a 5 minute cache, then a twentieth of the input price per read on Opus and Sonnet 5.5. Both run batch jobs at half the standard price.
Here is the same workload with half the input cached, and in batch:
| Model | List price | 50 in 100 input tokens cached | Batch price |
|---|---|---|---|
| GPT-5.5 | $220 | $175 | $110 |
| Claude Opus 5.5 | $160 | $122 | $80.00 |
| GPT-6.1 Sol | $80.00 | $61.00 | $40.00 |
| Claude Sonnet 5.5 | $80.00 | $61.00 | $40.00 |
| GPT-6.1 Terra | $40.00 | $30.50 | $20.00 |
| GPT-5 mini | $13.00 | $10.75 | $6.50 |
| Claude Haiku 5.5 | $4.00 | $3.10 | $2.00 |
| GPT-6 Luna | $4.00 | $3.10 | $2.00 |
Claude cached rows leave out cache writes, so treat them as a floor.
The fine print
- Several GPT models charge more once a prompt passes 272K tokens. Haiku 5.5 charges more above 100K.
- Anthropic notes that Claude 4.7 and later count more tokens for the same text than older models, so the price per token is not the whole story.
- OpenAI regional processing adds 10 in 100 to the price for models released from March 5, 2026.
You do not have to choose one
Many teams keep both keys: one provider for most work and the other where it does better. Ways to lower each bill are in cut your OpenAI API costs and cut your Anthropic API costs.