Free inference
Free tokens for your coding agent. On open models.
DeepSeek V4.1 Flash, GLM 5.3 Flash, GLM 5.3, gpt-oss 120B and Qwen3.8 Flash, free with your GitHub account, or with a card on file if you signed up with Google. Your dashboard shows how much you've used.
With GitHub, no credit card is needed: one allowance per GitHub account at least 30 days old. If you signed up with Google, you need a card on file to use free models. Free usage comes first, and anything beyond it is charged to your card by automatic top-up.
Free models
- DeepSeek V4.1 Flash
- GLM 5.3 Flash
- GLM 5.3
- gpt-oss 120B
- Qwen3.8 Flash
The models
5 open models, free to start. Then half the list price on shared capacity.
The same models on the paid API. Past your free usage, a balance pays for them: half the list price on shared capacity, and the list price when shared capacity is busy and another server takes the request.
- DeepSeek V4.1 Flash
deepseek-ai/DeepSeek-V4.1-FlashFree, then - GLM 5.3 Flash
zai-org/GLM-5.3-FlashFree, then - GLM 5.3
zai-org/GLM-5.3Free, then - gpt-oss 120B
openai/gpt-oss-120bFree, then - Qwen3.8 Flash
Qwen/Qwen3.8-FlashFree, then
Free usage
Limited, and it resets. Your dashboard shows how much is left.
Free usage is counted by the week, which resets every Monday. When it runs out, free requests stop until it resets.
- Your API keys page shows how much of it you have used, and when it resets.
Every request counts: the prompt, cached tokens included, and the reply. Longer requests and larger models use more of it.
The free usage is the workspace's: every key and member of it shares it.
If you signed up with Google, you need a card on file to use free models. Free usage comes first, and anything beyond it is charged to your card by automatic top-up. One allowance per card, and the card has to be a credit or debit card, not a prepaid one.
Past your free usage, a workspace with credit carries on at half the list price on shared capacity, and the list price when it's busy. Without credit, requests stop until it resets, and the answer says when (
free_allowance_used).When free capacity is busy, a workspace with credit is served instead at half the list price on shared capacity, and the list price when it's busy. Without credit, the answer is an overload error with a short
Retry-After(overloaded), which SDKs and agents retry by themselves.- Free requests are served by third-party model providers, who handle prompts and outputs under their own terms. Read the privacy policy.
Set up your agent
One line. Your agent does the rest.
Paste it into your coding agent. It sets itself up with DeepSeek V4.1 Flash and asks you for an API key: sign in with GitHub first, and the free tokens apply to it.
Set yourself up to use Runcrate's DeepSeek V4.1 Flash: run curl -fsSL "https://runcrate.ai/agents.md?model=deepseek-ai%2FDeepSeek-V4.1-Flash&intent=agent" and follow it.
Your agent asks for an API key. Create a key