Code
T7109 request-too-large
A request holds more than its model takes, and is refused before it is sent: the status 400. Some servers cut a long request with no word, so it is held here.
Details
The size is counted by the provider where it can count, and otherwise by a rule: one token for every 3 bytes of the text as UTF-8, and 1,500 for an image. What a model takes is in the list of its provider, or in takes of its definition. In a run with a conversation it is told as T7101, whose fix names shrink.
The fix
Send less: a shorter text, fewer files, or a shorter conversation. A model that takes more states it in takes.
TypeScript
export const local = model('local', { provider: 'openai-style', name: 'qwen3', url: 'http://localhost:11434/v1', abilities: ['reply'], takes: 128000 })