🧠Stop Using Opus for Typos
Pick the Right Model in Cursor
24 models. Different strengths. Use Opus for hard problems. Use Haiku for autocomplete. Stop mixing them up.
(Updated Dec 2025. Yes, we know GPT-5.1 isn't officially released yet. But it's in Cursor. 🤷)
🎯 Max Mode & Auto (The Cheat Codes)
- Max Mode: Gives larger context window for models like GPT-4.1, Gemini 2.5 Flash, and Gemini 3 Pro. Use for large refactors. Slower and pricier. Worth it.
- Auto: Cursor picks the best premium model automatically based on task and reliability. Use when you're lazy (like me) or when models are acting weird.
đź’ˇ Pro tip: Max Mode is not "better AI". It's just more context. Like giving GPT a longer memory. Don't use it for small tasks.
Filter models
Combine provider, type, capability, search, use-case
⚡ TL;DR picks (Because you don't read docs)
🏆 Daily driver
Claude 3.7 Sonnet or GPT-4o
Balance of quality + speed. Just use these.
🧠Hard problems
o1 or Claude 3 Opus
Deep reasoning only. Slow AF but worth it.
⚡ Speed runs
Claude 3.5 Haiku or Gemini 2.0 Flash
Blazing fast. Use for autocomplete and small edits.
Don't overthink it. Sonnet for daily work. Opus when stuck. Haiku for speed. Done. 🎯
Claude 3.7 Sonnet
Anthropic
Balanced
Best for
- Daily coding
- Long reasoning
- Refactors with context
Strengths
- Great code understanding
- Strong multi-file reasoning
- Fast enough for flow
Claude 3.5 Haiku
Anthropic
Speed
Best for
- Fast iterations
- Inline edits
- Lightweight chat
Strengths
- Very fast
- Low cost usage
- Good code summarization
Claude 3 Opus
Anthropic
Deep
Best for
- Deep design reviews
- Complex migrations
Strengths
- Top-tier reasoning
- Handles long context well
Slower and more expensive—use when depth matters
GPT-4o
OpenAI
Balanced
Best for
- Balanced daily driver
- UI/UX copy + code
- General chat + code
Strengths
- Strong coding & reasoning
- Good speed
- Multimodal capable
GPT-4o mini
OpenAI
Speed
Best for
- Quick edits
- Scaffolding
- Cheap test-gen
Strengths
- Fast and affordable
- Solid autocomplete + chat
GPT-4 Turbo
OpenAI
Deep
Best for
- Heavy refactors
- API design
- System prompts
Strengths
- Robust reasoning
- Good reliability on structured tasks
GPT-4.1
OpenAI
BalancedMax Mode
Best for
- General coding + chat
- Large-context work (Max Mode)
Strengths
- Strong reasoning
- Large context option in Max Mode
GPT-5.1 Codex
OpenAI
Balanced
Best for
- General coding + reasoning
- Mixed autocomplete + chat
Strengths
- Next-gen coding quality
- Stronger multi-line edits
Codex
OpenAI
Autocomplete
Best for
- General code completion
- Light chat about code
Strengths
- Reliable autocomplete
- Stable coding responses
Codex Fast
OpenAI
Speed
Best for
- Ultra-fast inline completion
- Short edits
- Tight feedback loops
Strengths
- Very fast latency
- Good for local iteration
Use larger models for long-form reasoning
Codex Max
OpenAI
Deep
Best for
- Deeper rewrites
- Bigger code suggestions
- Context-aware refactors
Strengths
- Stronger reasoning than Fast
- Better on multi-line edits
Slower than Codex Fast—reserve for bigger changes
o1
OpenAI
Deep
Best for
- Hard reasoning tasks
- Algorithmic problems
- Edge-case hunts
Strengths
- Excellent chain-of-thought
- Good for debugging corner cases
Slower; reserve for tricky reasoning, not bulk editing
Gemini 2.0 Pro
Balanced
Best for
- Structured data transforms
- Semantic search tasks
Strengths
- Good retrieval-style reasoning
- Solid code synthesis
Gemini 2.0 Flash
Speed
Best for
- Ultra-fast feedback
- Short completions
- Previewing snippets
Strengths
- Very fast responses
- Great for tight edit loops
Use Sonnet/GPT-4o for long-form reasoning
Gemini 2.5 Flash
SpeedMax Mode
Best for
- Max Mode large context
- Ultra-fast drafts with long files
Strengths
- Very large context in Max Mode
- Fast iterations
Gemini 3 Pro
BalancedMax Mode
Best for
- General coding with large context
- Design + reasoning
Strengths
- Strong reasoning
- Large context support
Composer 1
Composer
Balanced
Best for
- General coding assistance
- Inline improvements
Strengths
- Stable suggestions
- Good everyday throughput
DeepSeek V3
DeepSeek
Deep
Best for
- Math-heavy code
- Code translation
- Performance tuning ideas
Strengths
- Strong numeric reasoning
- Good at translation across languages
DeepSeek R1
DeepSeek
Balanced
Best for
- General-purpose coding
- Refactors with hints
Strengths
- Stable coding output
- Balanced speed vs quality
Opus 4.5
Anthropic
Deep
Best for
- Hard reasoning
- Large refactors
- Design reviews
Strengths
- High reasoning depth
- Great long-context handling
Use when depth outweighs speed
Sonnet 4.5
Anthropic
Balanced
Best for
- Daily driver with quality
- Multi-file edits
Strengths
- Balanced speed + quality
- Good code understanding
Grok 4
xAI
DeepMax Mode
Best for
- Researchy coding tasks
- Exploratory reasoning
Strengths
- Strong reasoning
- Good for exploration
Use when you need depth over speed
Cursor Small
Cursor
Autocomplete
Best for
- Small edits
- Low-latency autocomplete
- On-device friendly
Strengths
- Lightweight and fast
- Great for micro-changes
Use larger models for multi-file reasoning