Glass supports multiple language model providers, giving you flexibility in choosing the best models for your workflow. This guide covers setup and configuration for each provider.
Supported Providers
Anthropic Claude 4.6 Sonnet, Opus, and other Claude models
OpenAI GPT-4, GPT-5, o1, o3, and other OpenAI models
Ollama Local models for privacy and offline usage
OpenAI-Compatible Any OpenAI-compatible API endpoint
Copilot GitHub Copilot Chat integration
Google Gemini Gemini models via Agent Servers
Anthropic Claude Set up Anthropic’s Claude models:
Get API Key
Generate Key
Navigate to Settings → API Keys → Create Key
Copy Key
Save your API key (starts with sk-ant-)
Open Settings
cmd-, or ctrl-,
Navigate to AI
Settings → AI → Providers → Anthropic
Paste Key
Enter your API key and save
Set via environment variable: export ANTHROPIC_API_KEY = "sk-ant-..."
Or configure custom endpoint: { "language_models" : { "anthropic" : { "api_url" : "https://api.anthropic.com" } } }
Available Models Model Context Best For claude-4.6-sonnet 200k General use, balanced performance claude-4.6-opus 200k Complex reasoning, highest quality claude-4.6-sonnet-thinking 200k Tasks requiring reasoning steps
Model Context Notes claude-3.7-sonnet 200k Previous generation claude-3-opus 200k Previous generation claude-3-haiku 200k Fast, cost-effective
Custom Models Add custom Anthropic models:
{ "language_models" : { "anthropic" : { "available_models" : [ { "name" : "claude-custom-model" , "display_name" : "Claude Custom" , "max_tokens" : 200000 , "max_output_tokens" : 8192 , "cache_configuration" : { "max_cache_anchors" : 4 , "should_speculate" : true , "min_total_token" : 1024 } } ] } } }
OpenAI Configure OpenAI models:
Get API Key
Add Payment
Add payment method (required for API access)
Generate Key
Navigate to API Keys → Create new secret key
Copy Key
Save your API key (starts with sk-)
OpenAI API access requires a paid account with credits. GitHub Copilot subscription does not include API access.
Settings UI
Environment
JSON Config
Open Settings
cmd-, or ctrl-,
Navigate to AI
Settings → AI → Providers → OpenAI
Paste Key
Enter your API key and save
Set environment variable: export OPENAI_API_KEY = "sk-..."
Then restart Glass. { "language_models" : { "openai" : { "api_url" : "https://api.openai.com/v1" } } }
Available Models Model Context Best For gpt-5 200k Latest, best quality gpt-5-mini 200k Fast, cost-effective o3 128k Advanced reasoning o3-mini 128k Fast reasoning
Model Context Notes gpt-4o 128k Fast, multimodal gpt-4o-mini 128k Fastest, cheapest gpt-4-turbo 128k Previous generation
Custom Models { "language_models" : { "openai" : { "available_models" : [ { "name" : "gpt-4-custom" , "display_name" : "GPT-4 Custom" , "max_tokens" : 128000 , "max_output_tokens" : 4096 , "capabilities" : { "chat_completions" : true } } ] } } }
Ollama Run models locally with Ollama:
Setup Ollama
Install Ollama
Download from ollama.ai : # macOS brew install ollama # Linux curl -fsSL https://ollama.ai/install.sh | sh # Windows # Download installer from ollama.ai
Start Ollama
Or use system service (Linux/macOS):
Pull Models
Download code models: # For agents ollama pull qwen2.5-coder:32b # For completions ollama pull codellama:7b # For chat ollama pull llama3.3:70b
Configure Glass
{ "language_models" : { "ollama" : { "api_url" : "http://localhost:11434" } }, "agent" : { "default_model" : { "provider" : "ollama" , "model" : "qwen2.5-coder:32b" } } }
Recommended Models Model Size RAM Quality qwen2.5-coder:32b 19GB 32GB Best deepseek-coder-v2:16b 9GB 16GB Great codellama:13b 7GB 12GB Good mistral:7b 4GB 8GB Fast
Model Size Speed Quality codellama:7b 4GB Fast Good deepseek-coder:6.7b 4GB Fast Better starcoder:7b 4GB Fast Good codellama:13b 7GB Medium Better
Model Size Best For llama3.3:70b 40GB Best quality qwen2.5:32b 19GB Balanced mistral:7b 4GB Fast
Model Selection Ollama models appear in the model picker:
Only models that are currently pulled will appear in Glass.
OpenAI-Compatible APIs Connect to any OpenAI-compatible endpoint:
Configuration { "language_models" : { "openai_compatible" : { "api_url" : "https://your-endpoint.com/v1" , "api_key" : "your-key" , "available_models" : [ { "provider" : "custom" , "name" : "your-model" , "max_tokens" : 8192 } ] } } }
Compatible Services
LM Studio Local inference with UI: { "api_url" : "http://localhost:1234/v1" }
Together AI Hosted inference: { "api_url" : "https://api.together.xyz/v1" , "api_key" : "your-together-key" }
Azure OpenAI Azure deployment: { "api_url" : "https://your-resource.openai.azure.com/openai/deployments/your-deployment/chat/completions?api-version=2024-02-15-preview" , "api_key" : "your-azure-key" }
Replicate Model hosting: { "api_url" : "https://api.replicate.com/v1" , "api_key" : "your-replicate-token" }
GitHub Copilot Use GitHub Copilot Chat in Glass:
Requirements
Active GitHub Copilot subscription
GitHub account
Setup
Sign In
cmd-shift-p → “Sign in to GitHub”
Authorize Copilot
Follow the GitHub authentication flow
Enable
Copilot Chat is automatically enabled after sign-in
GitHub Copilot uses its own completions system separate from Glass AI features.
Google Gemini Access Gemini via Agent Servers:
Configuration { "agent_servers" : { "gemini" : { "command" : "npx" , "args" : [ "-y" , "@modelcontextprotocol/server-gemini" ], "env" : { "GEMINI_API_KEY" : "your-gemini-key" } } } }
Get API Key
Generate Key
Navigate to Get API Key → Create API Key
Configure
Add key to agent server config
Provider Selection Choose the right provider for your needs:
Decision Matrix Priority Recommended Provider Why Best quality Anthropic Claude Industry-leading models Latest models OpenAI Cutting-edge capabilities Privacy Ollama Local, offline, private Cost Ollama Free, unlimited Speed OpenAI GPT-4o Fastest cloud models Customization OpenAI-compatible Self-hosted, flexible
Multi-Provider Setup Use multiple providers simultaneously:
{ "agent" : { "default_model" : { "provider" : "anthropic" , "model" : "claude-4.6-sonnet" }, "inline_assistant_model" : { "provider" : "openai" , "model" : "gpt-4o-mini" }, "thread_summary_model" : { "provider" : "ollama" , "model" : "llama3.3:70b" } }, "edit_predictions" : { "provider" : "ollama" , "ollama" : { "model" : "codellama:7b" } } }
Troubleshooting
Verify key is correct (check for typos)
Ensure key has not expired
Check account has credits/subscription
Try regenerating the key
Verify API URL is correct
Check provider is authenticated
Verify API endpoint is reachable
Ensure provider is enabled in settings
Try refreshing Glass (cmd-r / ctrl-r)
Verify Ollama is running: ollama list
Check API URL: http://localhost:11434
Ensure models are pulled
Check firewall/network settings
Try restarting Ollama service
Check your API usage/quota
Wait and retry
Consider upgrading plan
Switch to different provider temporarily
Next Steps