<!-- https://zerolog.ch/open-source-models -->
Open-weight models

# Open-source models for your coding agent

ZeroLog's model picker lists open-weight models served by Featherless. You choose per conversation, see the price and context length of each, and pay with prepaid credit or your own provider key.

Request access

[See pricing](https://zerolog.ch/pricing)

Closed preview. Accounts are opened on request.

~/project Example

```
# one conversation, three models
1  low-cost model: explore the repository
2  strong model: the hard step
3  low-cost model: tidy up
```

## What is in the catalogue

- **Tens of thousands of models**

  The picker starts with a short recommended list. Switch to all Featherless models to search the full catalogue. The Models page adds filters for class and minimum context, sorting by newest, price, context or name, and tags: chat, tools, vision, reasoning, code and long context.
- **Models in the terminal**

  Claude Code and Codex can be pointed at the model you picked, for example Kimi K3, GLM or DeepSeek, with no Anthropic or OpenAI account involved. The catalogue changes: providers retire models, so check the picker for what is current.
- **Every row shows two numbers**

  Price per million tokens and context length, so you compare before you choose. Switch model mid-conversation and the thread comes along.
- **Open weights, not always open source**

  The catalogue is open-weight models. Licences differ from model to model, and ZeroLog does not check them for you. Read the licence of a model before you build a product on it.

## How you pay

- **Credit**

  Credit is spent at the provider's rate plus a 2 % service margin. The price under a reply already includes it, so the figure you see is the figure you are charged. At zero, paid models stop and nothing is billed afterwards.
- **Included and bought**

  Paid plans add credit every month, and included credit expires after 3 months. Packs can be bought on any plan, Free included. Prices are on the pricing page.
- **Your own key**

  Add a key under Settings, Provider keys, and the charge lands on your account at that provider instead of on ZeroLog credit, with no margin. After saving, only the last four characters stay visible. Cost is still shown per answer.

Details: [Choosing a model](https://zerolog.ch/docs/models), [Credit and top-ups](https://zerolog.ch/docs/credit), [Your own API key](https://zerolog.ch/docs/byok), [Cost per answer](https://zerolog.ch/docs/cost).

## What it does not offer

The picker shows only models that Featherless serves, so it has no Anthropic or OpenAI models. To use those, sign in to the vendor's CLI with your own account, as described on the Claude Code and Codex pages. Prompts sent to a catalogue model go to Featherless in the USA, and the chat shows the retention status for each answer. If a chosen model cannot answer, the chat shows an error and you pick another one.

## Pick a model and run an agent

1. **Open the model picker above the input.**

   Start with the recommended list. Compare price and context length.
2. **Search the full catalogue if you need to.**

   Filter by tag, class or minimum context.
3. **Start a CLI with that model.**

   In the container, Claude Code or Codex runs on the model you picked.
4. **Switch when the step changes.**

   A low-cost model to explore, a strong one for the hard step, a low-cost one to tidy up.

## Questions

- **Which open models can I use?**

  Those in the picker. It lists the open-weight models Featherless serves, and the list changes when providers retire or add models. Kimi K3, GLM and DeepSeek are examples named on the start page.
- **How much does a model cost?**

  The picker shows the price per million tokens for each model, and the price under a reply is what you are charged. It includes the 2 % margin. With your own key there is no margin.
- **Can I use my own API key?**

  Yes, under Settings, Provider keys. The charge then lands with your provider. The full key is hidden after saving.
- **Are the models really open source?**

  They have open weights. Licences vary per model, so read the licence before you build on one.
- **Which models can run inside Claude Code and Codex?**

  A model from the ZeroLog catalogue, or your own vendor account. Device tools such as the Android app are not available to the CLI engines.

## Keep reading

- **[Swiss AI workspace](https://zerolog.ch/swiss-ai-workspace)**

  Chat with many models and a Linux container in one account.
- **[Use your own Claude or ChatGPT subscription](https://zerolog.ch/bring-your-own-ai-subscription)**

  Claude Code and Codex on the plan you already pay for.
- **[Cloud development environment in Switzerland](https://zerolog.ch/swiss-cloud-dev-environment)**

  Where your files, your container and the model calls live.
- **[Claude Code in the cloud](https://zerolog.ch/claude-code-cloud)**

  Your own Claude account, signed in inside a container.
- **[Codex in a persistent container](https://zerolog.ch/codex-container)**

  OpenAI's Codex CLI with your own account.
- **[AI coding agent with a persistent workspace](https://zerolog.ch/ai-coding-agent-workspace)**

  Files, snapshots, jobs and schedules that outlast a chat.
- **[ZeroLog Code](https://zerolog.ch/code)**

  Your workspace, in your terminal.
- **[ZeroLog Extension](https://zerolog.ch/extension)**

  Your chat in the side panel of Chrome and Firefox.

## Try a different model for each step

Ask for access and compare models on your own code.

Request access

[See pricing](https://zerolog.ch/pricing)
