What is Chutes
Chutes provides public model inference and self-serve private deployments on confidential GPU infrastructure, with USD or TAO billing.
Pay as you go
As of September 17, 2026, the live catalogue lists 14 non-retired hosted models and public inference is billed per million input and output tokens at each model's live rate. Current examples include DeepSeek V4 Flash at US$0.44 input and US$1.32 output, Qwen 3.8 27B at US$0.24 and US$2.20, Kimi K3 at US$3 and US$15, GLM 5.1 at US$0.98 and US$3.08, and GLM 5.2 at US$1.25 and US$3.95.
Plus, Pro and Enterprise
Base costs US$3 per month with 300 bundled requests per day. Plus costs US$10 per month with 2,000 bundled requests per day and 6% off pay-as-you-go rates beyond the quota. Pro costs US$20 per month with 5,000 bundled requests per day and a 10% overage discount. Enterprise uses custom pricing for volume discounts, custom rate limits and dedicated support.
Private Chutes
Private deployments are billed by the second. The current verified confidential-GPU listing starts at US$1.80 per hour for an RTX Pro 6000, with a one-time deployment fee of three times the hourly rate, currently US$5.40 at that tier. Private Chutes support custom code, private weights, fine-tunes and dedicated production instances.
Official source: Chutes pricing.