API model catalogue

The table below lists the language models available inside SoftDesign. Rates are expressed in USD per token so you can estimate spend quickly. Multiply by 1,000 for per‑thousand token pricing. The default workspace model is gpt-5-nano.

Model Rate (USD/token) Approx. per 1K tokens
GPT-5.1 None 0.000200 0.2000
GPT-5.1 Low 0.000200 0.2000
GPT-5.1 Medium 0.000200 0.2000
GPT-5.1 High 0.000200 0.2000
GPT-5 0.000180 0.1800
GPT-5 mini 0.000082 0.0820
GPT-5 nano Default 0.000041 0.0410

Compute time estimates

Structural solves consume backend CPU/GPU time. SoftDesign currently estimates compute at 0.000066 USD per second, derived from mid‑range 8 vCPU cloud instances with a small overhead buffer. Multiply by your job's runtime (shown in usage history) to project the compute portion of your bill.

Rates are guidelines—you can override them via the OPENAI_TOKEN_RATE, MODEL_TOKEN_RATES, and COMPUTE_SECOND_RATE environment variables when deploying.