Model Selection

Learn to select the right AI model for each task. This 10-minute guide will help you balance speed, cost, and capability to maximize productivity while controlling expenses.

Model Landscape Overview

Tip

Key Principle: Start with Claude Opus 4.5 as your default. It’s the best coding model available (>80% SWE-Bench). Add specialized models like Composer 1 for speed or GPT-5.1-Codex-Max for bug fixing when needed.

Quick Reference Matrix

Model	Speed	Cost	Context	Best Use Case
Claude Opus 4.5	⚡⚡	$$$$$	200k	Default for all coding - best model (>80% SWE-Bench)
Cursor Composer 1	⚡⚡⚡⚡	$$	TBD	Speed-critical work (4x faster)
GPT-5.1-Codex-Max	⚡⚡⚡	$$	200k+	Bug fixing & UI generation
Claude Sonnet 4.5	⚡⚡⚡	$	1M	Budget-conscious, large context needs
Gemini 3 Pro	⚡⚡⚡	$$	1M	Extreme context & multimodal analysis

Cost Comparison

Monthly Budget Impact

Typical Developer Usage

• With Opus 4.5 (recommended): ~$100-200/month
• Budget with Sonnet 4.5: ~$50/month
• Optimized mix: ~$100-150/month

Token Economics

Per Million Tokens

• Opus 4.5: $5 (input) / $25 (output)
• Sonnet 4.5: $3 / $15
• GPT-5.1-Codex-Max: $1.25 / $10
• Gemini 3 Pro: $2 / $12
• Composer 1: Premium tier

Claude Opus 4.5 - The Default Choice

Released: November 24, 2025 (announcement)

When to Use Opus 4.5 (Default)

Perfect For

• All coding tasks (default for everything)
• Feature implementation
• Architecture design
• Code reviews
• Security audits
• Agent building and computer use
• Complex refactoring
• Test writing
• Documentation
• API integration
• Database queries

Why It's Best

• First to break 80% on SWE-Bench Verified
• Best coding model in the world
• Best at building complex agents and computer use
• Enhanced prompt injection resistance
• Memory improvements for sustained focus
• Effort parameter for adjustable reasoning depth
• Recommended with Max/Ultra subscription plans

Opus 4.5 Optimization Tips

1. Clear, specific prompts yield better results than vague requests
2. Leverage effort parameter for adjustable reasoning depth
3. Best for agentic workflows - let it handle multi-step tasks
4. Use project rules to maintain consistency
5. Get Max/Ultra plans for full access

Real-World Examples

// EXCELLENT: Leverage Opus 4.5's superior reasoning
"Analyze the entire authentication system across all files,
identify security vulnerabilities, and refactor to use
industry best practices. Include tests for all changes."

// GOOD: Specific request for Opus 4.5
"Create a TypeScript function that validates email addresses
using regex, returns a Result<string, ValidationError> type,
and includes unit tests"

Cursor Composer 1 - Speed Champion

Released: October 29, 2025 (announcement)

When to Use Composer 1

Decision Framework

Use Composer 1 when you need:

1. Speed: 250 tokens/sec (4x faster than other models)
2. Rapid Iteration: Quick feedback loops for iterative work
3. High Throughput: Multiple coding sessions in succession
4. Budget-Conscious Speed: Better than Sonnet 4.5 for speed/price in Cursor

Tips for Composer 1

1. Use for rapid iteration - great for quick changes
2. Better than Sonnet 4.5 in Cursor - faster and more cost-effective
3. Most turns complete in under 30 seconds
4. Trained with codebase-wide semantic search - understands large projects

Example Workflow

// Use Composer 1 for rapid iteration
"Make this button 10px larger and change its color to blue"

// Use Composer 1 for quick refactoring
"Extract this logic into a separate utility function"

// Switch to Opus 4.5 for complex architecture
"Design a new authentication system with OAuth support"

GPT-5.1-Codex-Max - Bug Fixing & UI Specialist

Released: November 19, 2025 (announcement) Available in: Cursor, GitHub Copilot

When to Use GPT-5.1-Codex-Max

Decision Framework

Use GPT-5.1-Codex-Max when you need:

1. Bug Fixing: Specialized training for identifying and fixing bugs quickly
2. UI Generation: Excellent at creating and refining user interfaces
3. Frontend Work: Strong understanding of React, Vue, Angular, and modern CSS
4. Long-Running Tasks: Compaction enables 24+ hour task endurance

Tips for GPT-5.1-Codex-Max

1. Use for targeted bug fixes - especially production issues
2. Great for component creation - UI components, forms, layouts
3. Combine with Opus 4.5 - use Codex-Max for bugs/UI, Opus for architecture
4. Leverage 24+ hour capability - for extended debugging sessions

Example Workflow

// Use GPT-5.1-Codex-Max for bug fixing
"This button isn't responding to clicks on mobile.
Debug and fix the touch event handlers."

// Use GPT-5.1-Codex-Max for UI generation
"Create a responsive navigation menu with dropdown support,
mobile hamburger menu, and smooth animations."

// Switch to Opus 4.5 for broader work
"Now integrate this navigation into our routing system
and add proper authentication checks."

Claude Sonnet 4.5 - Budget Alternative

Tip

When to Use: Sonnet 4.5 is a cost-effective alternative when Opus 4.5 is too expensive or when you need the larger 1M context window.

When to Use Sonnet 4.5

Decision Framework

Use Sonnet 4.5 when:

1. Budget-Conscious: Need to save costs on token usage
2. Large Context Needs: Tasks requiring >200K tokens (Opus 4.5’s limit)
3. Quota Exhausted: When your Opus 4.5 quota runs out

Note for Cursor users: Composer 1 is often a better choice than Sonnet 4.5 due to its 4x speed advantage.

Sonnet 4.5 vs Composer 1 in Cursor

Aspect	Sonnet 4.5	Composer 1
Speed	Standard	4x faster
Context	1M tokens	TBD
Best For	Large context needs	Speed-critical work
Recommendation	Use for >200K context	Use for most other cases

Gemini 3 Pro - Multimodal & Extreme Context

Released: November 18, 2025 (announcement)

Tip

When to Use: Best multimodal model with 1M context. Use for image/video analysis or when you exceed Opus 4.5’s 200K context.

When to Use Gemini 3 Pro

Use Gemini 3 Pro when:

• Tasks require multimodal analysis (images, video, diagrams)
• Your context exceeds Opus 4.5’s 200K limit
• Need Deep Think mode for complex reasoning
• Analyzing visual documentation or screenshots

Key Features

• 1M token context - analyze very large codebases
• 1501 Elo on LMArena - top ranking
• Deep Think mode - for complex reasoning
• Best multimodal - images, video, audio analysis

Model Selection Decision Tree

graph TD Start[New Task] --> Type{Task Type?} Type -->|Bug Fix or UI| Codex[GPT-5.1-Codex-Max] Type -->|Speed Critical| Composer[Cursor Composer 1] Type -->|Everything Else| Opus[Claude Opus 4.5 - Default] Codex --> Working{Working well?} Composer --> Working3{Working well?} Opus --> Working2{Working well?} Working -->|Yes| Done[Continue] Working -->|No| Opus Working3 -->|Yes| Done Working3 -->|No| Opus Working2 -->|Yes| Done Working2 -->|No| Check{Context > 200K?} Check -->|Yes| Gemini[Gemini 3 Pro] Check -->|No| Budget{Budget constrained?} Budget -->|Yes| Sonnet[Sonnet 4.5] Budget -->|No| Stay[Stay with Opus 4.5]

Cost Optimization Strategies

1. Progressive Enhancement

1. Start with Opus 4.5 as the default (best model)
2. Use Composer 1 for speed-critical work in Cursor
3. Use GPT-5.1-Codex-Max for bug fixes and UI generation
4. Use Sonnet 4.5 when budget-conscious or need >200K context
5. Use Gemini 3 Pro for multimodal or extreme context needs

2. Context Management

Reduce Context

• Clear chat regularly
• Use focused @mentions
• Exclude irrelevant files
• Summarize long discussions

Reuse Context

• Save useful prompts
• Create project rules
• Build prompt templates
• Use memory feature

3. Batch Operations

# EFFICIENT: Single comprehensive call with Opus 4.5
"Refactor auth, user, and profile services to follow
our new architecture pattern. Provide implementation
plan first, then execute."

# SPEED-FOCUSED: Use Composer 1 for quick iterations
"Make these three quick changes to the button styles"

Model Switching Workflow

Quick Switch Method

Tip

Use keyboard shortcut Ctrl + . to quickly switch models mid-conversation

Strategic Switching Pattern

1. All Phases (Opus 4.5 - Default)

   • Exploration, planning, implementation
   • Architecture design
   • Complex problem solving
   • Code review

2. Speed-Critical Phases (Composer 1)

   • Rapid iterations
   • Quick changes
   • High-throughput sessions

3. Specialized Phases (GPT-5.1-Codex-Max)

   • Bug fixing
   • UI generation
   • Frontend work

4. Large Context Phases (Gemini 3 Pro or Sonnet 4.5)

   • Analyzing >200K tokens
   • Full codebase analysis

Real-World Scenarios

Scenario 1: New Feature

Task: Add user notifications
Models Used:
  - Opus 4.5: Full implementation (100%)
Total Cost: ~$5
Time Saved: 4 hours

Scenario 2: UI-Heavy Feature

Task: Build dashboard with charts
Models Used:
  - Opus 4.5: Architecture (30%)
  - GPT-5.1-Codex-Max: UI components (70%)
Total Cost: ~$8
Time Saved: 6 hours

Scenario 3: Large Refactor

Task: Migrate to new framework
Models Used:
  - Opus 4.5: Planning and implementation (80%)
  - Gemini 3 Pro: Large codebase analysis (20%)
Total Cost: ~$30
Time Saved: 20 hours

Advanced Model Techniques

1. Model Chaining

Opus 4.5: "Create a detailed plan for implementing OAuth"
  ↓ (Save plan to file)
Opus 4.5: "Implement step 1 from oauth-plan.md"
Composer 1: "Quick fixes for step 2" (speed)
GPT-5.1-Codex-Max: "Create the login UI component"

2. Parallel Model Usage

Open multiple Cursor instances:

• Instance 1: Opus 4.5 for architecture
• Instance 2: Composer 1 for rapid iteration
• Instance 3: GPT-5.1-Codex-Max for UI work

3. Model-Specific Prompting

Opus 4.5 Prompts

- High-level goals
- Ask for reasoning
- Request alternatives
- Include constraints
- Leverage effort parameter

Composer 1 Prompts

- Direct and specific
- Quick iterations
- Reference file paths
- Clear success criteria

GPT-5.1-Codex-Max Prompts

- Bug-focused context
- Include error logs
- UI specifications
- Visual requirements

Monitoring Model Usage

Track Your Costs

1. Check usage: Settings → Usage
2. Set budget alerts
3. Review weekly patterns
4. Optimize based on data

Usage Patterns to Watch

• High token usage: Leverage Opus 4.5’s efficiency
• Repeated similar tasks: Create rules/templates
• Long conversations: Clear context more often
• Speed bottlenecks: Switch to Composer 1

Model Selection Checklist

Before starting a task, ask:

• Is this a standard coding task? (Use Opus 4.5 - default)
• Is this a bug fix or UI work? (Use GPT-5.1-Codex-Max)
• Need speed in Cursor? (Use Composer 1)
• Need >200K context? (Use Gemini 3 Pro or Sonnet 4.5)
• Am I using the right prompting style?
• Is the model working well or should I switch?

Next Steps

Continue to Project Rules

Now let’s set up project rules to ensure consistent AI behavior across all models.

Project Rules →

Time: 10 minutes

Tip

Key Takeaway: Claude Opus 4.5 is THE BEST coding model (>80% SWE-Bench) - use it as your default. Complement with Composer 1 for speed, GPT-5.1-Codex-Max for bug fixes and UI work. Get Max/Ultra plans for full access. Check Cursor changelog for updates.