Others calculate sticker price.
We calculate the bill you receive.
Your invoice isn't (input+output)×price. It's cache hit / miss / write priced in three parts, reasoning tokens billed separately, the whole request re-priced when prompt crosses a context tier, tool rounds resending results, and retries. Those fields are where llmburn lives.
Open the calculator → Methodology 国产模型中文站
Why trust the numbers
- Three independent sources — OpenRouter × LiteLLM (commit-pinned) × models.dev, cross-validated; LiteLLM's openrouter mirror keys don't count as evidence.
- 228 of 462 models reach multi-source agreement; 128 have disputes we show side by side, never averaged.
- 81 models publish long-context tier jumps — the #1 hidden cost no other calculator shows.
- Missing data is shown as 未验证, never filled with a plausible-looking default.
Flagship models — real-bill fields
| Model | input | output | cache read | cache write |
|---|---|---|---|---|
| anthropic-claude-opus-5-5 | $4–$4.80† | $20–$24† | $0.2–$0.24† | $5–$6† |
| openai-gpt-6-1-sol-pro | $2 | $10 | $0.1 | $2.50 |
| google-gemini-3-8-flash | $0.75 | $3.75 | $0.075 | $0.042 |
| x-ai-grok-4-7 | $2 | $6 | $0.5 | 未验证 |
| anthropic-claude-sonnet-5-5 | $2–$2.40† | $10–$12† | $0.2–$0.24† | $2.50–$3† |
| google-gemini-3-1-pro-preview | $2 | $12 | $0.2 | $0.375 |
| xiaomi-mimo-v2-6-pro-ultraspeed | $4.35 | $8.70 | $0.036 | 未验证 |
| xiaomi-mimo-v2-6-flash | $0.14 | $0.28 | $0.0028 | 未验证 |
| xiaomi-mimo-v2-6-pro | $0.435 | $0.87 | $0.0036 | 未验证 |
| qwen-qwen3-8-omni-flash | $0.15 | $0.47 | $0.016 | 未验证 |
| z-ai-glm-5-3-flashx | $0.37 | $1.25 | $0.075–$0.09† | $0 |
| qwen-qwen3-8-max-0902 | $2 | $6 | $0.25 | $2.50 |
† sources disagree — values shown side by side on the model page, never averaged.
All 302 models with real-bill fields →
Snapshot 2026-10-01 · Machine-readable: /data/v1/pricing.json · Estimates from public sticker prices, not your invoice.