Let talk about the fact that most ai chatbots have a hard time remembering things. You start vibe coding, something else comes up, you close your browser window, switch pc’s , etc.
Now there are memory settings and projects that allow you to preserve things, sometimes.
But what I have found is , that if you can, sit down and just get from start to finish in one session, things go pretty well.
Some AI Chatbots have limitations on the amount of chatting you can do, others not so much.
I asked Chat GPT to compare all 5 ai chatbots, the amount they allow you to chat for free.
Here is what it came up with:
compare the following 5 ai chatbots, chat gpt, copilot, gemini, claude and deepseek , how much can you chat with the free versions and what capabilities do the free versions have, any limitations
| AI | Chat limits (free) | Model access | Strengths | Biggest limitations |
|---|---|---|---|---|
| ChatGPT | Moderate–high (soft caps, slows under load) | Smaller GPT models | Balanced, tools, coding, general use | Limited advanced models/tools |
| Microsoft Copilot | Generally generous (depends on mode) | GPT-based (via Microsoft) | Web + Office integration | Less customizable, tied to MS ecosystem |
| Google Gemini | High (but daily caps on some features) | Gemini Flash/basic | Strong free features, Google integration | Some strict prompt caps in features |
| Claude | Low–moderate (tight rolling limits) | Claude Sonnet (free) | Excellent writing, long documents | Hits limits quickly |
| DeepSeek | Very high / near-unlimited | Full model access | Powerful + free reasoning | Less polished UI, fewer integrations |
From experience I can tell you , Chat GPT and Cluade work best, but even on the first paid tier I seem to run out of credits, regularly, deepseek seems to never run out of credits but is slow most of the time and frequently wrong on its first coding attempts.
I also asked about memory features:
compare their abilities to store previous conversations to reference in the future
| Feature | ChatGPT | Claude | Gemini | Copilot | DeepSeek |
|---|---|---|---|---|---|
| Remembers across chats | ✅ Yes | ✅ Yes | ⚠️ Limited | ❌ Mostly no | ❌ No |
| Learns preferences over time | ✅ Strong | ✅ Strong | ⚠️ Partial | ❌ Minimal | ❌ No |
| Automatic memory | ⚠️ Partial | ✅ Strong | ⚠️ Indirect | ❌ No | ❌ No |
| External data memory | ❌ Limited | ❌ Limited | ✅ Strong | ✅ Some | ❌ No |
| Editable memory | ✅ Yes | ✅ Yes | ⚠️ Some | ❌ No | ❌ No |
So each one has its strengths and weaknesses, Deepseek is great for long chats, Claude and Chat GPT are good for projects, so depending how big your coding project is, how many seesions it might take, might influence your choice.
My fallback is also the prompt to create the step by step setup guide with complete code, if you are needing to leave deepseek after hours of prompting, ask it for that, then reupload it for reference.
My Vibe Coding Projects 