Step-5-Preview
StepFunprovides a 1M-token context window
implementation detail · filed under model architecture
A context capacity of up to one million tokens, distinguished from the separate 256K context window.
The sources on this page only mention it in passing. The line above is the extraction's one-line gloss. See context capacity for the methods filed beside it.
Also called long context window, One-million-token context window.
5 quotes from 5 sources, one card per model, each opening with its strongest.
provides a 1M-token context window
1M-token context window, suited for mixed-modality input and multi-agent collaboration
1M context is now the default across all official DeepSeek services
Its 1M context window supports complete documents, extended conversations, and complex task contexts in a single pass
supports up to 1 million tokens of context
supports up to 1 million tokens of context