← Back to Docs

Llama Models

Meta's open-weight models for code generation and reasoning.

Llama 3.3 70B

ProviderMeta AI
Parameters70 billion
Context128K tokens
Best forCode generation, analysis, structured output
Dispatch roleImplementation, spec writing, documentation
InfrastructureDispatch Pro cloud
CostIncluded in Pro ($20/month, shared 5M token pool)

Llama 3.3 70B is a strong general-purpose model with excellent code generation capabilities. Dispatch routes medium-complexity tasks to Llama 3.3 when quality and speed need to be balanced -- implementation tasks, documentation generation, and structured analysis.

Llama 4 Scout 17B

ProviderMeta AI
Parameters17 billion (MoE)
Context128K tokens
Best forFast code tasks, triage, exploration
Dispatch roleQuick implementation, boilerplate, triage
InfrastructureDispatch Pro cloud
CostIncluded in Pro ($20/month, shared 5M token pool)

Llama 4 Scout is a fast, efficient mixture-of-experts model. With 17B active parameters, it responds quickly while maintaining strong code quality. Dispatch routes simpler tasks to Scout when speed matters more than reasoning depth -- config changes, boilerplate generation, quick fixes, and triage decisions.

Configuration

Available to all Dispatch Pro subscribers. Set any pipeline stage to llama-3.3-70b or llama-4-scout-17b in Settings > Pipeline Profiles, or let Auto Model route automatically based on task complexity.

Isolation

All cloud model runs use worktree isolation.