chutes.ai/

Chutes

chutes.ai/

Chutes provides serverless compute for deploying AI models in production — language, image, video, speech, and music models — describing itself as "the leading open-source, decentralized compute provider" so customers don't manage GPU hardware directly. Its hero claims to be "Powering Trillions of Tokens per Month," and the page features a live model marketplace listing 10+ LLMs (including recent releases like Kimi K2.6 and GLM 5.2) with real-time per-model pricing rather than a flat rate card. A Python code sample for calling the API sits high on the page, ahead of the three named pricing tiers (Plus, Pro, Enterprise), aimed squarely at a developer audience.

Visit website

Category

Engineering & Infrastructure

Style

Dark Colorful

Stack

Next.js Tailwind CSS

Fonts

Tomato Grotesk

Palette

Meta title

Chutes | Serverless AI Compute

Sections

Updated

Oct 10, 2026

Chutes's landing page, section by section

Each section below is cropped straight from Chutes's live page — useful references for engineering & infrastructure landing page design.

Chutes design review: how the homepage is built

The homepage runs from a dark hero with a colour-gradient cube render through a partner logo strip, a model section, a product grid, a pricing block, a searchable model index and a closing call to action, across 8,239 pixels of desktop height. The colour approach is near-black, with #070607 behind the sections, #fefefe for the primary white buttons and #2c29c9 as the violet of the model marks in the cards. The detected display typeface is Tomato Grotesk, and no body face was detected. The page brands itself as Chutes, matching meta.name.

Above the fold

The hero headline "Breakthrough Serverless Compute for AI, at Scale." uses a multicolour gradient on its first word only. Beneath it, a subhead describes Chutes as the leading open-source, decentralized compute provider for running open-source models in production. Two buttons follow, a white "Explore Models" and an outlined "Create Account", with "Chutes Global" and the "bittensor" wordmark below. The top bar reads "Product", "Resources", "Company" and "Pricing", with "Sign In" at the right.

Partner strip and LLM section

A strip of five partner wordmarks, "OpenRouter", "KILO", "fetch.ai", "Phala" and "Pax Historia", sits under the hero. The section headed "SOTA Open-Source LLMs, Available here first." offers three "JUST LANDED" chips, "Kimi K2.6", "GLM 5.2" and "Gemma 4 31B", plus a dark Python code panel labelled "HOW TO USE".

Model tiles, cube cards and feature rows

The "There's a Chute for Everything" grid opens with four category tiles, "Image", "Video", "Speech" and "Music", then five model cards priced at "$1.80/hr". Below, "Made for hyperscaling AI powered products" introduces three cube cards, "Purpose-built for AI Developers", "Designed to be flexible, but fast" and "Decentralized, Open-source Infrastructure". Icon features follow, headed "AI Model Inference", "TEE/Secure Compute" and "Consumer Apps".

Pricing tiers and model index

Pricing shows three tiers: "Plus" at "$10 per month", "Pro" at "$20 per month" with a "Best Value" band, and "Enterprise" with "Contact us". "Top Models Available Always On, Ready to Scale." follows with a "Search Chutes" box and a "Popularity" sort, then a grid of ten model cards such as "Qwen3 32B", "DeepSeek V3.2" and "Kimi K3".

Closing section and footer

The closing band reads "AI Compute for Everyone." above "Talk to sales" and a green "Get started" button. The footer has five columns, "Features", "Product", "Company", "Legal" and "Connect", and the base line shows "© 2026 Chutes. All rights reserved." with "Built by Chutes" at the right.

On mobile

At 390px the hero stacks its headline, subhead and two buttons vertically.Model cards drop to a two-column grid with truncated names, and the cube cards stack into one column. The three-up feature row becomes a two-plus-one arrangement. Pricing tiers stack, the "Pro" card keeps its "Best Value" band, and footer columns pair up in two columns.

Worth borrowing

The hero sets a gradient on one headline word and pairs two buttons, one white and one outlined, under a single subhead.
Model cards show a name, a slug, tags and a per-hour or per-token price, so the catalogue reads like a price list.
Three cube renders serve as card art for the product claims, pairing each short claim with one image.