Markdown pages, the full model catalogue and a default model

Markdown versions and llms.txt

The home page, pricing, every documentation topic and each Newsroom article now have a Markdown version. Add .md to the address, for example /pricing.md, or send Accept: text/markdown. Browsers still get the regular page.

/llms.txt is now built from the same data as the home and pricing pages. /llms-full.txt puts the home page, pricing and all docs into one file. /.well-known/api-catalog lists the public monitoring API.

Documentation and structured data

A new docs page, Isolation, explains how workspaces are kept apart from each other and from the host. Spend limits now says how they differ from machine and rate limits. Every public page carries one consistent set of structured data, with operator details matching the legal notice and prices taken from the pricing page.

Temporary model pause covers the current pause on Featherless models: your account, credit and history stay, and paid time is extended by the length of the pause.

The model catalogue and a default for new chats

A chat started without a chosen model used to fail. It now uses the default model, GLM-5.3-Flash, and your settings page shows which model that is instead of an internal placeholder.

We tested the Featherless models on ten fixed tasks, from arithmetic and code to tool calls, and measured speed, context length and price. The recommended tiers now follow those results: GLM-4.7-Flash for everyday use, GLM-5.3-Flash for most work, and Qwen3.8 2.4T for long, multi-step tasks. DeepSeek-V4-Flash left the everyday tier after failing four of the ten tasks.

The model list now comes from the complete Featherless catalogue and is refreshed every hour. Models from other providers are no longer offered. The picker opens on a short recommended list. Switch to All Featherless models to search everything, or open Models to filter by class and context and sort by newest, price or context.