Max Tokens
Domain: Models & Providers
Motivation
maxTokens is the per-request cap on generated tokens. On the DeepSeek adapter it defaults to 256,000; a model’s own cap and an explicit request value win over the default. The adapter materializes a configured output cap when a request omits one, so even uncapped requests stay bounded.
Concrete Example
config: { maxTokens: 256000 } on dsh-llm-deepseek-api-key; for purpose: 'session-title' requests the adapter forces thinking off to reserve output for visible title text.
Analogy
The word-count limit on an essay — you can write up to here and no more.