Understand All Available Models and Their Variant Configurations

What You'll Learn

• Choose the Claude or Gemini model that best fits your needs
• Understand different Thinking mode levels (low/max or minimal/low/medium/high)
• Understand the two separate quota pools: Antigravity and Gemini CLI
• Use the --variant parameter to dynamically adjust thinking budget

Your Current Challenge

You've just installed the plugin and are facing a long list of model names, unsure which to choose:

• What's the difference between antigravity-gemini-3-pro and gemini-3-pro-preview?
• What does --variant=max mean? What happens if you don't specify it?
• Are Claude's thinking mode and Gemini's thinking mode the same?

Core Concept

Antigravity Auth supports two main categories of models, each with an independent quota pool:

1. Antigravity quota: Accessed via Google Antigravity API, including Claude and Gemini 3
2. Gemini CLI quota: Accessed via Gemini CLI API, including Gemini 2.5 and Gemini 3 Preview

Variant System

OpenCode's variant system lets you avoid defining separate models for each thinking level. Instead, you specify the configuration at runtime using the --variant parameter. This makes model selection cleaner and configuration more flexible.

Antigravity Quota Models

These models are accessed with the antigravity- prefix and use the Antigravity API quota pool.

Gemini 3 Series

Gemini 3 Pro

Model Name	Variants	Thinking Levels	Description
antigravity-gemini-3-pro	low, high	low, high	Balances quality and speed

Variant Configuration Examples:

# Low thinking level (faster)
opencode run "Quick answer" --model=google/antigravity-gemini-3-pro --variant=low

# High thinking level (deeper)
opencode run "Complex reasoning" --model=google/antigravity-gemini-3-pro --variant=high

Gemini 3 Flash

Model Name	Variants	Thinking Levels	Description
antigravity-gemini-3-flash	minimal, low, medium, high	minimal, low, medium, high	Ultra-fast response, supports 4 thinking levels

Variant Configuration Examples:

# Minimal thinking (fastest)
opencode run "Simple task" --model=google/antigravity-gemini-3-flash --variant=minimal

# Balanced thinking (default)
opencode run "Regular task" --model=google/antigravity-gemini-3-flash --variant=medium

# Maximum thinking (deepest)
opencode run "Complex analysis" --model=google/antigravity-gemini-3-flash --variant=high

Gemini 3 Pro Doesn't Support Minimal/Medium

gemini-3-pro only supports low and high levels. If you try to use --variant=minimal or --variant=medium, the API will return an error.

Claude Series

Claude Sonnet 4.5 (Non-Thinking)

Model Name	Variants	Thinking Budget	Description
antigravity-claude-sonnet-4-5	—	—	Standard mode, no extended thinking

Usage Example:

# Standard mode
opencode run "Daily conversation" --model=google/antigravity-claude-sonnet-4-5

Claude Sonnet 4.5 Thinking

Model Name	Variants	Thinking Budget (tokens)	Description
antigravity-claude-sonnet-4-5-thinking	low, max	8192 (low) / 32768 (max)	Balanced mode

Variant Configuration Examples:

# Light thinking (faster)
opencode run "Quick reasoning" --model=google/antigravity-claude-sonnet-4-5-thinking --variant=low

# Maximum thinking (deepest)
opencode run "Deep analysis" --model=google/antigravity-claude-sonnet-4-5-thinking --variant=max

Claude Opus 4.5 Thinking

Model Name	Variants	Thinking Budget (tokens)	Description
antigravity-claude-opus-4-5-thinking	low, max	8192 (low) / 32768 (max)	Strongest reasoning capability

Variant Configuration Examples:

# Light thinking
opencode run "High-quality answer" --model=google/antigravity-claude-opus-4-5-thinking --variant=low

# Maximum thinking (for most complex tasks)
opencode run "Expert-level analysis" --model=google/antigravity-claude-opus-4-5-thinking --variant=max

Claude vs Gemini Thinking Mode Differences

• Claude uses a numeric thinking budget (tokens), such as 8192, 32768
• Gemini 3 uses string-based thinking levels (minimal/low/medium/high)
• Both display their reasoning process before responding, but configuration methods differ

Gemini CLI Quota Models

These models don't have the antigravity- prefix and use the Gemini CLI API's separate quota pool. They don't support thinking mode.

Model Name	Description
gemini-2.5-flash	Gemini 2.5 Flash (fast response)
gemini-2.5-pro	Gemini 2.5 Pro (balances quality and speed)
gemini-3-flash-preview	Gemini 3 Flash Preview (preview version)
gemini-3-pro-preview	Gemini 3 Pro Preview (preview version)

Usage Examples:

# Gemini 2.5 Pro (no thinking)
opencode run "Quick task" --model=google/gemini-2.5-pro

# Gemini 3 Pro Preview (no thinking)
opencode run "Preview model testing" --model=google/gemini-3-pro-preview

Preview Models

gemini-3-*-preview models are official Google preview versions and may be unstable or subject to change. If you want Thinking capabilities, use the antigravity-gemini-3-* models.

Model Comparison Overview

Feature	Claude 4.5	Gemini 3	Gemini 2.5
Thinking Support	✅ (thinkingBudget)	✅ (thinkingLevel)	❌
Google Search	❌	✅	✅
Quota Pool	Antigravity	Antigravity + Gemini CLI	Gemini CLI
Use Cases	Complex reasoning, coding	General tasks + search	Fast response, simple tasks

🎯 How to Choose Models

Choose Claude or Gemini?

• Choose Claude: You need stronger logical reasoning, more stable code generation
• Choose Gemini 3: You need Google Search, faster response speeds

Choose Thinking or Standard Mode?

• Use Thinking: Complex reasoning, multi-step tasks, need to see reasoning process
• Use Standard Mode: Simple Q&A, fast response, no reasoning display needed

Which Thinking Level?

Level	Claude (tokens)	Gemini 3	Use Cases
minimal	—	Flash only	Ultra-fast tasks like translation, summarization
low	8192	Pro/Flash	Balances quality and speed, suitable for most tasks
medium	—	Flash only	Moderately complex tasks
high/max	32768	Pro/Flash	Most complex tasks like system design, deep analysis

Recommended Configurations

• Daily development: antigravity-claude-sonnet-4-5-thinking --variant=low
• Complex reasoning: antigravity-claude-opus-4-5-thinking --variant=max
• Quick Q&A + search: antigravity-gemini-3-flash --variant=low + Google Search enabled

Complete Configuration Example

Add the following configuration to ~/.config/opencode/opencode.json:

{
  "$schema": "https://opencode.ai/config.json",
  "plugin": ["opencode-antigravity-auth@latest"],
  "provider": {
    "google": {
      "models": {
        "antigravity-gemini-3-pro": {
          "name": "Gemini 3 Pro (Antigravity)",
          "limit": { "context": 1048576, "output": 65535 },
          "modalities": { "input": ["text", "image", "pdf"], "output": ["text"] },
          "variants": {
            "low": { "thinkingLevel": "low" },
            "high": { "thinkingLevel": "high" }
          }
        },
        "antigravity-gemini-3-flash": {
          "name": "Gemini 3 Flash (Antigravity)",
          "limit": { "context": 1048576, "output": 65536 },
          "modalities": { "input": ["text", "image", "pdf"], "output": ["text"] },
          "variants": {
            "minimal": { "thinkingLevel": "minimal" },
            "low": { "thinkingLevel": "low" },
            "medium": { "thinkingLevel": "medium" },
            "high": { "thinkingLevel": "high" }
          }
        },
        "antigravity-claude-sonnet-4-5": {
          "name": "Claude Sonnet 4.5 (Antigravity)",
          "limit": { "context": 200000, "output": 64000 },
          "modalities": { "input": ["text", "image", "pdf"], "output": ["text"] }
        },
        "antigravity-claude-sonnet-4-5-thinking": {
          "name": "Claude Sonnet 4.5 Thinking (Antigravity)",
          "limit": { "context": 200000, "output": 64000 },
          "modalities": { "input": ["text", "image", "pdf"], "output": ["text"] },
          "variants": {
            "low": { "thinkingConfig": { "thinkingBudget": 8192 } },
            "max": { "thinkingConfig": { "thinkingBudget": 32768 } }
          }
        },
        "antigravity-claude-opus-4-5-thinking": {
          "name": "Claude Opus 4.5 Thinking (Antigravity)",
          "limit": { "context": 200000, "output": 64000 },
          "modalities": { "input": ["text", "image", "pdf"], "output": ["text"] },
          "variants": {
            "low": { "thinkingConfig": { "thinkingBudget": 8192 } },
            "max": { "thinkingConfig": { "thinkingBudget": 32768 } }
          }
        },
        "gemini-2.5-flash": {
          "name": "Gemini 2.5 Flash (Gemini CLI)",
          "limit": { "context": 1048576, "output": 65536 },
          "modalities": { "input": ["text", "image", "pdf"], "output": ["text"] }
        },
        "gemini-2.5-pro": {
          "name": "Gemini 2.5 Pro (Gemini CLI)",
          "limit": { "context": 1048576, "output": 65536 },
          "modalities": { "input": ["text", "image", "pdf"], "output": ["text"] }
        },
        "gemini-3-flash-preview": {
          "name": "Gemini 3 Flash Preview (Gemini CLI)",
          "limit": { "context": 1048576, "output": 65536 },
          "modalities": { "input": ["text", "image", "pdf"], "output": ["text"] }
        },
        "gemini-3-pro-preview": {
          "name": "Gemini 3 Pro Preview (Gemini CLI)",
          "limit": { "context": 1048576, "output": 65535 },
          "modalities": { "input": ["text", "image", "pdf"], "output": ["text"] }
        }
      }
    }
  }
}

Copy Configuration

Click the copy button in the top-right corner of the code block above, then paste it into your ~/.config/opencode/opencode.json file.

Checkpoint ✅

Complete these steps to confirm you've mastered model selection:

• [ ] Understand the two separate quota pools: Antigravity and Gemini CLI
• [ ] Know that Claude uses thinkingBudget (tokens), Gemini 3 uses thinkingLevel (strings)
• [ ] Can choose the appropriate variant based on task complexity
• [ ] Have added the complete configuration to opencode.json

Lesson Summary

Antigravity Auth provides rich model selection and flexible variant configurations:

• Antigravity quota: Supports Claude 4.5 and Gemini 3 with Thinking capabilities
• Gemini CLI quota: Supports Gemini 2.5 and Gemini 3 Preview without Thinking capabilities
• Variant system: Dynamically adjust thinking levels via the --variant parameter without defining multiple models

When choosing models, consider your task type (reasoning vs search), complexity (simple vs complex), and response speed requirements.

Coming Up Next

In the next lesson, we'll learn Thinking Models Explained.

You'll learn:

• The principles behind Claude and Gemini Thinking modes
• How to configure custom thinking budgets
• Techniques for preserving thinking blocks (signature caching)

---

Appendix: Source Code Reference

Click to view source code locations

Updated: 2026-01-23

Feature	File Path	Line Numbers
Model parsing and tier extraction	src/plugin/transform/model-resolver.ts	177-282
Thinking tier budget definitions	src/plugin/transform/model-resolver.ts	14-19
Gemini 3 thinking level definitions	src/plugin/transform/model-resolver.ts	26
Model alias mapping	src/plugin/transform/model-resolver.ts	36-57
Variant configuration parsing	src/plugin/transform/model-resolver.ts	374-422
Type definitions	src/plugin/transform/types.ts	1-115

Key Constants:

• THINKING_TIER_BUDGETS: Mapping of thinking budgets for Claude and Gemini 2.5 (low/medium/high → tokens)
• GEMINI_3_THINKING_LEVELS: Supported thinking levels for Gemini 3 (minimal/low/medium/high)

Key Functions:

• resolveModelWithTier(requestedModel): Resolves model name and thinking configuration
• resolveModelWithVariant(requestedModel, variantConfig): Resolves model from variant configuration
• budgetToGemini3Level(budget): Maps token budget to Gemini 3 levels