Skip to main content

Earn Credits

Share your device's idle compute with the Teale network and earn Teale Credits. You earn credits continuously while online and additional credits when your device actually serves a request. Credits redeem AI inference on any model, and you can gift them to other users.

note

Credits are a closed-loop unit — they redeem AI inference and transfer to other users, but they do not cash out to fiat or stablecoins. See Credit Economy for the full model.

Prerequisites​

Quick start​

teale up --maximize-earnings

This single command starts Teale in full earning mode. It handles model selection, networking, and power management automatically.

What --maximize-earnings does​

When you pass --maximize-earnings, Teale makes several optimizations:

  1. Keeps the system awake. Prevents sleep so your node stays available to the network around the clock.
  2. Auto-manages models. Downloads and loads the largest model that fits in your available RAM (reserving 4 GB for macOS). If network demand shifts, Teale may swap models automatically.
  3. Increases storage allocation. Raises the default model storage limit so Teale can cache multiple models for fast switching.
  4. Enables WAN networking. Connects to the Teale relay so requests can reach you from the public network, not just your LAN.

Model selection by RAM​

Teale picks the largest quantized model that fits in your available memory (total RAM minus 4 GB for the OS):

Mac RAMUsable for inferenceTypical model
16 GB~12 GB8B (4-bit)
32 GB~28 GB32B (4-bit)
64 GB~60 GB70B (4-bit)
128 GB~124 GB70B (8-bit) or 405B (4-bit)

Larger models are in higher demand and earn more per token.

Monitor your earnings​

Check your balance at any time:

teale wallet balance

View recent transactions:

teale wallet transactions --limit 10

Check node status (loaded model, active connections, requests served):

teale status

Optimization tips​

  1. Plug into AC power. Teale throttles inference on battery to preserve battery health. AC power unlocks full throughput.

  2. Use ethernet. Wired connections reduce latency and improve reliability, which means your node wins more routing decisions.

  3. Schedule overnight contribution. If you prefer not to share compute during the day, configure a contribution schedule:

    teale config set schedule.earn "22:00-08:00"

    Or in the desktop app: Settings > Contribution Schedule.

  4. Keep Teale updated. Newer versions include performance improvements and support for newer models.

  5. More RAM = more earnings. Larger models serve more use cases and command higher prices.

How pricing works​

Providers earn 95% of the inference cost. The remaining 5% is a network fee that funds relay infrastructure and protocol development.

The cost of a request is calculated as:

cost = (tokens / 1000) * (params * 0.1) * quantMultiplier / 10000
  • tokens --- number of tokens generated
  • params --- model parameter count in billions (e.g. 8 for an 8B model)
  • quantMultiplier --- adjusts for quantization level (higher precision = higher cost)

Electricity cost floor​

Teale ensures you never lose money by enforcing an electricity cost floor. If the market price for a request would be less than the electricity required to serve it, your node declines the request.

Configure your local electricity rate:

teale config set electricity_cost 0.12

The value is in USD per kWh. Teale uses this along with measured power draw to calculate your cost floor.

Configure earning without --maximize-earnings​

You can enable earning mode with finer control:

teale up
teale config set wan_enabled true
teale config set auto_manage_models true
teale config set max_storage_gb 50

This gives you the same behavior with explicit control over each setting.

Next steps​