Request too large or too long

If a request has too much text for the model, or the request body is too big, the API refuses it. Shorten the conversation, lower max_tokens, or send large documents in a smarter way. These errors do not go away on retry; the request has to change.

Codes you may see

Status	Code	Message	Limit
400	context_length_exceeded	”The request exceeds the model’s maximum context length. Shorten the messages or lower max_tokens.”	Input plus max_tokens must fit the model’s context window (262,144 tokens for most models).
413	payload_too_large	”The request is too large.”	A request body of about 2 MB.
413	too_many_messages	”The request has too many messages.”	500 messages per request.

Fixes

• Lower max_tokens. The space reserved for the answer counts toward the window, so a large max_tokens can push a request over.
• Trim old turns. Summarize or drop the oldest messages of a long conversation.
• Count before you send. Use token counting and trim when you are close to the limit.
• Send documents as files and quote only the relevant parts. See Files.
• For agents, turn on context management so older content is cleared automatically.

In the apps

App	What to do
Web app	Start a new conversation for a new topic.
Syntic Desktop	”This conversation is too long. Start a new chat or upgrade.” Start a new chat.
Syntic Code	Type /compact to summarize older messages, or /clear to start fresh. /context shows what fills the conversation. Syntic Code also compacts automatically when the conversation gets full.

Related

• Syntic Code usage analytics and limits
• Error codes reference