OpenCode Models - Configure Claude, GPT, Gemini

Supported AI Models

OpenCode supports multiple AI providers and models, giving you flexibility to choose the best model for your coding tasks. Each model has unique strengths for different use cases.

Anthropic Claude Models

Claude models excel at code understanding, complex reasoning, and following detailed instructions.

Model	Best For	Context Window
claude-sonnet-4-20250514	Daily coding tasks, balanced performance	200K tokens
claude-opus-4-20250514	Complex architecture, deep analysis	200K tokens
claude-3-5-haiku	Quick tasks, high speed	200K tokens

Configure Claude:

bash

# Set API key
export ANTHROPIC_API_KEY="sk-ant-..."

# Use in OpenCode
opencode --model claude-sonnet-4-20250514

OpenAI GPT Models

GPT models are widely used and offer strong general-purpose coding capabilities.

Model	Best For	Context Window
gpt-4o	Complex reasoning, multi-step tasks	128K tokens
gpt-4o-mini	Fast responses, simple tasks	128K tokens
o1	Advanced reasoning, planning	128K tokens

Configure GPT:

bash

# Set API key
export OPENAI_API_KEY="sk-..."

# Use in OpenCode
opencode --model gpt-4o

Google Gemini Models

Gemini models offer excellent multimodal capabilities and large context windows.

Model	Best For	Context Window
gemini-2.0-flash	Fast, efficient coding	1M tokens
gemini-1.5-pro	Large codebase analysis	2M tokens

Configure Gemini:

bash

# Set API key
export GOOGLE_API_KEY="AIza..."

# Use in OpenCode
opencode --model gemini-2.0-flash

Choosing the Right Model

For Daily Development

Recommended: Claude Sonnet or GPT-4o-mini

These models offer the best balance of speed, cost, and capability for everyday coding tasks like:

Writing new functions
Debugging code
Code reviews
Documentation

For Complex Projects

Recommended: Claude Opus or GPT-4o

Use these powerful models when working on:

System architecture design
Large-scale refactoring
Complex algorithm implementation
Security audits

For Large Codebases

Recommended: Gemini 1.5 Pro

With its massive 2M token context window, Gemini excels at:

Analyzing entire repositories
Understanding complex dependencies
Cross-file refactoring

Configuration File

Set your default model in ~/.config/opencode/config.yaml:

bash

# Default model configuration
model: claude-sonnet-4-20250514

# Provider-specific settings
providers:
  anthropic:
    api_key: ${ANTHROPIC_API_KEY}
    default_model: claude-sonnet-4-20250514

  openai:
    api_key: ${OPENAI_API_KEY}
    default_model: gpt-4o

  google:
    api_key: ${GOOGLE_API_KEY}
    default_model: gemini-2.0-flash

# Model aliases for quick switching
aliases:
  fast: claude-3-5-haiku
  smart: claude-opus-4-20250514
  large: gemini-1.5-pro

Switching Models

Command Line

bash

# Use specific model for a session
opencode --model gpt-4o

# Use alias
opencode --model fast

During Session

Switch models without restarting:

bash

/model claude-opus-4-20250514

Or use the model picker:

bash

/model

Local Models

OpenCode also supports local models through Ollama:

bash

# Install Ollama
curl -fsSL https://ollama.com/install.sh | sh

# Pull a model
ollama pull codellama:13b

# Configure in OpenCode
opencode --model ollama:codellama:13b

Recommended Local Models

Model	Size	Best For
codellama:13b	7GB	Code completion, simple tasks
deepseek-coder:33b	19GB	Complex coding, analysis
qwen2.5-coder:32b	18GB	Multi-language support

Performance Tips

Match model to task: Use faster models for simple tasks, powerful models for complex ones
Use aliases: Set up shortcuts for frequently used models
Monitor token usage: Larger contexts cost more with cloud models
Consider latency: Local models eliminate network latency but require GPU resources

Cost Optimization

Strategy	Description
Start small	Begin with cheaper models, escalate when needed
Use caching	Enable response caching to reduce API calls
Batch requests	Group related questions to maximize context usage
Local fallback	Use local models for simple tasks

Troubleshooting

Model Not Available

bash

Error: Model 'claude-opus-4' not found

Solution: Check the exact model name in provider documentation.

Rate Limiting

bash

Error: Rate limit exceeded

Solution: Configure rate limiting in config:

bash

rate_limits:
  anthropic:
    requests_per_minute: 50
    tokens_per_minute: 100000

Context Length Exceeded

bash

Error: Context length exceeded

Solution: Use a model with larger context window or enable automatic context management:

bash

context:
  auto_truncate: true
  max_tokens: 100000