As we continue to innovate and push the boundaries of AI capabilities, we’re often asked about our approach to context length. At SambaNova, we take a pragmatic and iterative approach to expanding context lengths for our models. We initially release models with a context length that balances performance, efficiency, and usability, and then incrementally increase it as demand and use cases evolve.
For instance, our DeepSeek-R1 model has seen its context length grow to 16k tokens, while DeepSeek-V3-0324 is at 8k and is slated to reach 16k soon. We’re committed to providing the flexibility and scalability that our users need, and we’re pleased to see our models being used in increasingly complex and demanding applications.
As a reminder, our context lengths represent the total token capacity, encompassing both input tokens (including system prompts) and output tokens. You can stay up-to-date on the latest context lengths for our models by visiting our documentation at Supported models - SambaNova Documentation.
We’re dedicated to continuing to advance the state-of-the-art in AI, and we’re committed to working closely with our community to understand their needs and priorities. While we can’t provide a one-size-fits-all solution for context length, we’re focused on delivering the right balance of performance, efficiency, and usability for a wide range of use cases.
-Coby