Llama 3.2 3B Instruct
Llama 3.2 3B Instruct is available through the each::labs LLM Router. The router handles authentication, retries, and provider fallback so you can target this model with a single API surface, no per-provider SDK required.
- Router slug
- meta-llama/llama-3.2-3b-instruct
- Provider
- Meta-llama
- Model id
- llama-3.2-3b-instruct
- Context
- 131K tokens
- Input
- text
- Price / M tokens
- $0.05 in · $0.34 out
- Web search
- Not supported
Call it from cURL
curl -X POST \
-H "Authorization: Bearer $EACHLABS_API_KEY" \
-H "Content-Type: application/json" \
--data '{
"model": "eachlabs-llm-router",
"input": {
"model": "meta-llama/llama-3.2-3b-instruct",
"messages": [
{ "role": "user", "content": "Hello!" }
]
},
"webhook_url": ""
}' \
https://api.eachlabs.ai/v1/prediction/Why route?
- One key, one endpoint, swap models without rotating SDKs.
- Automatic retries and provider failover when Meta-llama hits rate limits.
- Unified billing through each::labs, pay-as-you-go.
- Drop-in observability, every call is traced in the dashboard.
Working with Llama 3.2 3B Instruct
About Llama 3.2 3B Instruct
Llama 3.2 3B Instruct is a large language model from Meta-llama, available through the each::labs LLM Router. You call it with a single API surface, the router handles authentication, retries, and provider failover so your code stays portable.
Quick start
1. Grab your each::labs API key from the dashboard. 2. POST to https://api.eachlabs.ai/v1/prediction/ with an Authorization: Bearer header and a body of { "model": "eachlabs-llm-router", "input": { ... } }. 3. Pass meta-llama/llama-3.2-3b-instruct as the model field inside input, alongside your messages array. 4. The create call returns a predictionID; poll GET /v1/prediction/{predictionID} until status is "success".
The cURL snippet above is ready to copy.
Why route through each::labs
- One API, many models. Switch from Llama 3.2 3B Instruct to any other LLM by changing one string, no SDK rotation.
- Automatic failover. Configure a backup model and the router fails over the moment Meta-llama returns an error or times out.
- Unified observability. Every call lands in your each::labs traces with latency, token usage, and cost broken out per request.
- One invoice. Pay as you go in USD, at each model's published price.
Production notes
- The router is the same path each::labs uses internally, no preview gates.
- Streaming, system prompts, and tool calls are passed through to Meta-llama as-is.
- Rate limits follow Meta-llama's upstream limits; the router queues and retries on transient 429s automatically.
About Llama 3.2 3B Instruct
Llama 3.2 3B Instruct is a large language model from Meta-llama, exposed through the eachlabs-llm-router. You target it with a single API and a single key, no Meta-llama-specific SDK required.
Send a request to the eachlabs-llm-router endpoint with meta-llama/llama-3.2-3b-instruct as the model input. The cURL and TypeScript snippets above show the full call.
One key, one endpoint, one bill. The router handles authentication, retries, automatic failover when Meta-llama rate limits or times out, and traces every call in your each::labs dashboard.
Yes. The router supports cross provider failover, list backup models in your call config and we'll switch the moment Meta-llama returns an error.
Pay as you go through your each::labs invoice. Token level usage is recorded against the provider's published price; there's no platform markup.
Yes. The router is the same path each::labs uses internally, no preview gates, no waitlists. Once your each::labs account is active, you can call Llama 3.2 3B Instruct from production immediately.