alt Hacker NewsHow does the context length scaling at 256K tokens compare to Llama's 1M in terms of performance? How are the contexts treated differently?
How does the context length scaling at 256K tokens compare to Llama's 1M in terms of performance? How are the contexts treated differently?