Pricing
Per-token model pricing
Prepaid credit is consumed per token at the rates below. We pass on our volume discounts — you pay 20% less than going directly to each provider.
| Model | Input per 1M | Output per 1M | Cache read / write |
|---|---|---|---|
| Claude Sonnet 4 Anthropic | $2.40 $3.00 | $12.00 $15.00 | R $0.24 W $3.00 |
| GPT-4.1 OpenAI | $1.60 $2.00 | $6.40 $8.00 | $0.40 |
| Gemini 2.5 Pro Google | $1.00 $1.25 | $4.00 $5.00 | — |
| DeepSeek V3 DeepSeek | $0.22 $0.27 | $0.88 $1.10 | R $0.06 W $0.22 |
| Llama 3.3 70B Meta (self-hosted) | $0.47 $0.59 | $0.63 $0.79 | — |
| Qwen 2.5 Coder Alibaba | $0.40 $0.50 | $1.20 $1.50 | $0.12 |
Input
Tokens sent to the model — your prompt, code context, and instructions.
Output
Tokens generated by the model — code, explanations, and diffs.
Cache read
Reusing cached context is significantly cheaper than re-sending it.
Cache write
First-time context storage. Some providers charge separately, others unify read/write.
Prices are per 1 million tokens (1M = 1,000,000). Rates are rounded to the nearest cent. Credits never expire. BuildStax rates are 20% below provider list price.