api providers

get your free api key and start renaming layers

— cerebras (recommended)

cerebras is the default provider because it offers:

• free api access — no credit card required
• blazing fast inference — 3,000+ tokens per second
• powerful models — including gpt-oss-120b with 128k context window

1. create a cerebras account

   go to cloud.cerebras.ai and sign up for free.

2. generate an api key

   once logged in, navigate to the api keys section and create a new key. copy it.

3. paste in layersense

   open layersense settings → api tab → paste your key in the “api key” field.

4. select a model

   recommended models:

   • gpt-oss-120b — best quality, 128k context, 3,000+ tok/s
   • llama-3.3-70b — balanced performance, 2,100 tok/s
   • qwen-3-32b — fast and efficient, 2,600 tok/s

cerebras free tier has generous rate limits. if you hit them, the plugin automatically retries with delays.

— google gemini (free)

google offers a generous free tier for gemini models.

1. get your api key

   go to aistudio.google.com and create an api key.

2. configure layersense

   settings → api tab → provider: google gemini → paste your key.

3. select a model

   • gemini-3-pro — newest, most intelligent (preview)
   • gemini-2.5-flash — best price-performance, 1M context
   • gemini-2.5-pro — advanced thinking model

— openrouter

openrouter aggregates multiple providers and offers some free models.

1. create an account

   go to openrouter.ai and sign up.

2. get your api key

   find your api key in the dashboard.

3. configure layersense

   settings → api tab → provider: openrouter → paste your key.

4. use free models

   look for models tagged :free like:

   • mistralai/devstral-2512:free — 123b coding specialist
   • meta-llama/llama-3-8b-instruct:free — balanced all-purpose

— local (lm studio)

run models on your own computer for complete privacy and no rate limits.

1. download lm studio

   get it from lmstudio.ai — available for windows, mac, and linux.

2. download a model

   inside lm studio, search for and download a model. recommended:

   • qwen 2.5 (various sizes)
   • llama 3 8b
   • mistral 7b
3. start the local server

   in lm studio, go to the “local server” tab and click start. the default endpoint is http://127.0.0.1:1234/v1/chat/completions.

4. configure layersense

   settings → api tab → provider: local (lm studio) → verify the endpoint matches your lm studio server.

local models require a decent gpu for fast inference. cpu-only works but is slower.

— paid providers

if you already have api keys from paid providers:

provider	setup	recommended models
openai	paste your key, select provider “openai”	gpt-5.1 (best for agents), gpt-5-nano (cheapest), gpt-4.1-mini
anthropic	paste your key, select provider “anthropic”	claude-sonnet-4-5 (best balance), claude-haiku-4-5 (fastest), claude-opus-4-5 (most capable)

— which provider to choose?

your situation	recommended provider
just want it to work for free	cerebras with gpt-oss-120b (3,000+ tok/s)
need maximum privacy	local with lm studio
already have openai/anthropic keys	use your existing provider
want cutting-edge intelligence	google gemini with gemini-3-pro
want to try different models	openrouter for variety