01
Why usage runs out
ChatGPT plans have usage limits that depend on the model, the tools you use, and how long your conversations are. Bigger models, deep reasoning, agents, and image or video generation use more.
The fastest way to hit a limit is one long chat on the biggest model.
02
Know your plan
- Free and Go
- Lower limits. Good for lighter use.
- Plus
- $20 a month, with Codex and higher limits.
- Pro
- $100, $200, or $500 a month for heavier use.
- Business
- Per-seat plans with admin controls.
Limits change often. Check your plan’s current limits in ChatGPT’s settings and help center.
03
Pick the right model
- GPT-6 Luna
- Focused, repeatable tasks. Uses the least.
- GPT-6.1 Sol
- Complex work and agentic tasks.
- GPT-6 Astra
- The hardest problems. Uses the most.
Start with the smaller model. Move up only when the answer is not good enough.
04
Use less reasoning effort
- Lower effort for quick edits and simple questions
- Higher effort for planning, analysis, and debugging
- Deep research modes only for real research projects
05
Start fresh chats
- One chat per topic
- Summarize, then continue in a new chat
- Avoid reusing old chats for new tasks
Long chats carry all their history into every new message.
06
Write complete prompts
- Give context, goal, and format in one message
- Ask related questions together
- Ask for the length you need
- Edit your message instead of sending corrections
07
Use projects and memory
Put files and instructions you reuse in a project. Let memory keep your preferences, so you do not repeat them in every chat.
08
Turn off tools you do not need
Web search, connected apps, and agent features add work to each request. Use them when the task needs them, not by default.
09
Generate images and video efficiently
- Write a detailed prompt the first time
- Edit an image instead of regenerating it
- Make drafts before final high-quality versions
10
Save Codex usage
- Use GPT-6 Luna for focused coding tasks
- Lower reasoning effort for routine edits
- Keep AGENTS.md short
- Start new chats for new tasks
- Check usage and limits with /status
11
Use agents carefully
Agents and always-on Dots can use a lot of capacity. Give them narrow tasks and stop runs that go the wrong way.
12
When you hit a limit
- Switch to a smaller model
- Finish simple tasks while you wait
- Use another tool for light work
- Upgrade only if limits block real work every week
13
Master usage-saving prompt
Answer briefly with no preamble. Use bullet points and keep it under [length].
If anything is unclear, ask one question instead of guessing.
Context: [context]. Task: [task]. Format: [format].
The golden rule
Do not use the biggest model for everything. Match the model, the effort, and the chat length to the task.
- Model
- Effort
- Fresh chats
- Projects
- Tools
- Codex
For your next five tasks, start with the smaller model. Move up only when you must.