AI Rate CardCoding agent cost calculator
What a coding agent costs a month on each model: tasks a day, model calls per task, the context each call sends, its answer and reasoning, and the share read from the prompt cache - at the cheapest real-time seller of each model. How to use it
9 000 model calls a month: 270 000 000 input and 20 700 000 output tokens.
| Model | Reasoning class | Provider | One task | A month |
|---|---|---|---|---|
| DeepSeek V4.1 Flash | 🧠 A | openrouter | $0.0283 | $8.50 |
| Claude Haiku 5.5 | 🧠 B | abacus | $0.0678 | $20.34 |
| GLM-5.3 | 🧠 A | edenai | $0.11 | $32.85 |
| Qwen3-Coder 480B-A35B Instruct | trustedrouter | $0.18 | $55.45 | |
| Gemini 3.8 Flash | 🧠 A | kenari | $0.22 | $66.89 |
| Kimi K3 | 🧠 A | aihubmix | $0.27 | $81.51 |
| Qwen3.8 Max | 🧠 A | edenai | $0.72 | $216.74 |
| Gemini 3.1 Pro Preview | 🧠 A | kilo | $0.75 | $224.10 |
| Grok 4.6 | 🧠 B | edenai | $1.08 | $324.00 |
| Claude Sonnet 5.5 | 🧠 S | abacus | $1.29 | $387.90 |
| GPT-5.5 | 🧠 A | xpersona | $2.18 | $653.40 |
| Claude Opus 5.5 | 🧠 S | abacus | $2.59 | $775.80 |
| GPT-6 Astra | 🧠 A | meganova | $5.42 | $1627.20 |
Formula a month: tasks × calls × 30 × (context × input price, the cached share at the cache-read price + (answer + reasoning) × output price). An agent resends its growing context with every call, so input usually dominates. Not counted: cache write fees, tool and search fees, failed retries.