
Open-source benchmark for evaluating coding-agent harnesses on Terminal-Bench tasks inside hardened, NIST-derived sandbox policies. Supports Inspect AI, Harbor, Daytona sandboxes, and OpenRouter model inference.
Prices and medians update for the tier you select.
Ranked by how closely each one matches boundarybench's job. Prices show each provider's Pro state; entry prices are labelled as such. Unpriced products still belong to the market.
Market = the products most similar to this one by capability; prices are median / quartiles over its priced members, separated by provider type and buyer tier.