Every OpenAI model
Each model on its default cache setting. Select a model for its worked examples, break-even rule and caching rules.
| Model | Input | 30 min write | 12 hour write | Cached input | Output | Min. cacheable | Coding agent | Without | Share saved |
|---|---|---|---|---|---|---|---|---|---|
| GPT-6 Astra | $10.00 | $12.50 | $25.00 | $1.00 | $50.00 | 1,024 | $6.81 | $38.48 | 82% |
| GPT-6.1 Sol | $2.00 | $2.50 | $5.00 | $0.10 | $10.00 | 1,024 | $1.00 | $7.70 | 87% |
| GPT-6.1 Terra | $1.00 | $1.25 | $2.50 | $0.05 | $5.00 | 1,024 | $0.50 | $3.85 | 87% |
| GPT-6 Sol | $2.00 | $2.50 | $5.00 | $0.20 | $10.00 | 1,024 | $1.36 | $7.70 | 82% |
| GPT-6 Luna | $0.10 | $0.125 | $0.25 | $0.01 | $0.50 | 1,024 | $0.068 | $0.38 | 82% |
| GPT-5.6 Sol | $4.00 | $5.00 | $10.00 | $0.40 | $20.00 | 1,024 | $2.72 | $15.39 | 82% |
| GPT-5.6 Terra | $2.00 | $2.50 | $5.00 | $0.20 | $12.00 | 1,024 | $1.41 | $7.74 | 82% |
| GPT-5.6 Luna | $0.20 | $0.25 | $0.50 | $0.02 | $1.20 | 1,024 | $0.14 | $0.77 | 82% |
| GPT-5.5 | $5.00 | None | None | $0.50 | $30.00 | Varies | $3.50 | $19.36 | 82% |
| GPT-5.4 | $2.50 | None | None | $0.25 | $15.00 | Varies | $1.66 | $9.68 | 83% |
| GPT-5.4 mini | $0.75 | None | None | $0.075 | $4.50 | Varies | $0.71 | $2.90 | 75% |
Coding agent: 40 turns, 20K tokens of starting context, 3K in and 600 out per turn, 1.5 minutes apart, three long breaks of 8 minutes.
How OpenAI caching works
- Caching is automatic. GPT-5.6 and later charge for writes, at 30 minute or 12 hour minimum lifetimes, and need at least 1,024 tokens.
- Older models have no write charge. OpenAI gives typical lifetimes for them rather than fixed ones, so the simulator uses the low end and says so.
- Above 272K input tokens, long-context prices apply. Assumed: they apply to the whole request, cached and written tokens included.
Sources: pricing page and caching guide. Each model page lists the rules that apply to it.
Other providers
Anthropic Google All 32 models compared
Try any of these sessions with your own numbers in the simulator. It runs in your browser and shows where the cache goes cold.