13. Providers, Models and Model Selection

13.1 Provider vs model

A provider supplies access to AI models.

A model is the actual model used for inference.

Conceptually:

OpenCode
   |
   +--> Provider A --> Model A
   |
   +--> Provider B --> Model B
   |
   +--> Provider C --> Model C

13.2 Multiple providers

OpenCode's current documentation states that it supports many LLM providers and local models.

Provider availability changes over time.

Use:

/connect

inside OpenCode to connect a provider.

Use the current /models interface and official provider documentation to inspect available models.

13.3 Model IDs

OpenCode configuration uses provider/model-style identifiers in its documented agent configuration.

Example:

provider/model-id

Always copy the exact model identifier exposed by your provider/OpenCode installation.

13.4 Model selection

Think about:

Task
 |
 +--> Required reasoning
 +--> Coding ability
 +--> Speed
 +--> Cost
 +--> Context needs
 |
 v
Model choice

Example strategy:

This is a task-based framework, not a ranking of providers or models.

13.5 Agent-specific models

OpenCode supports configuring a model for an agent.

Example:

{
  "agent": {
    "plan": {
      "model": "provider/model-id"
    },
    "review": {
      "model": "provider/model-id"
    }
  }
}

Use the exact configuration schema for your installed OpenCode version.

13.6 Local models

OpenCode documentation also describes support for local models.

Local models can be useful when:

• Data must remain local
• You want to experiment
• You have suitable hardware
• Latency/cost requirements favor local inference

Local models may differ significantly in capability from hosted models.

Exercise

For five tasks in your project, define:

1. Task
2. Required capability
3. Preferred model characteristics
4. Why
5. Verification method