You see Yikes! Info Overload or Response Ran Out of Tokens. This means the chat holds more content than the AI model can read at once.
What causes it
Large videos with long transcripts
Many PDFs or documents connected to one chat
A long conversation history
Each model has a token limit (how much text it can read at once):
Claude 5 Sonnet: 200k
Claude Opus 5.5, Opus 5, Opus 4.8, Opus 4.6: 1 million
GPT-5.6 Terra, GPT-5.5, Gemini 3.8 Flash, DeepSeek V4.1 Flash, DeepSeek V4 Pro: 1 million
GPT-5.4 Mini: 400k
Grok 4.7, Grok 4.5: 500k
See the AI Models Guide.
Fixes
Switch to a model with a bigger limit
Click the model name under the text box and pick Gemini 3.8 Flash, GPT-5.6 Terra or GPT-5.5. Claude Opus models also have 1 million, if your plan includes them.
Connect less
Disconnect files you don't need for this question.
Use separate chats for separate topics.
Connect large files to a separate chat, ask it to extract the key points, put the result in a text box (press "T") and connect that to your main chat.
If the chat is too good to lose: move it to a new chat
Switch to a model with a 1 million limit.
Send: "Summarize all the key learnings and context from our conversation so another chat can pick up exactly where we left off"
Copy the summary into a new chat, or into a text box and connect it.
These fixes also lower your credit usage. See Troubleshooting High Credit Usage.