This model requires you to accept the DeepSeek license terms before download.
Description
DeepSeek-MoE 16B chat — fused-MoE runtime, greedy decode contract
Tags
At a glance
| Kind | language model |
|---|---|
| Precision | bfloat16 |
| Container size | 30.5 GiB |
| Weights inside | 30.5 GiB |
| Components | 2 |
| Licence | DeepSeek |
What is inside the container
A .nbx container packs every part of deepseek-moe-16b-chat — 2 components with weights, graphs and runtime settings — so one file is all the runtime needs.
| Component | Blocks | Layers | Weights |
|---|---|---|---|
| model | 0 | 28 | 30.1 GiB |
| lm_head | 0 | — | 400 MiB |
Default settings
The values the container ships with. They apply when you do not pass your own.
- Max length
- 4096
- Temperature
- 0
- Top-p
- 1
- Top-k
- 50
Memory needed
The weights of deepseek-moe-16b-chat take 30.5 GiB in bfloat16. That is the floor for the memory the runtime has to place, spread over the devices it finds; activations come on top and grow with the size and length you ask for. A measured peak per device class is not published for this model yet.
Verify what you downloaded
Every download carries its SHA-256, and the file endpoint sends it in the X-NBX-SHA256 header. The container is 30.5 GiB (32,755,132,988 bytes).
$ sha256sum deepseek-moe-16b-chat.nbx
849ff339da88ccea370ad1c76df74224d33c8503acca61cd8612c12bfefa8947
Questions
What is deepseek-moe-16b-chat?
A language model by deepseek-ai, distributed as a bfloat16 .nbx container and licensed DeepSeek.
How big is the download?
30.5 GiB, of which 30.5 GiB are weights.
How do I run it?
Install the runtime from the download page, then use the commands under “CLI Usage” below: neurobrix import deepseek-ai/deepseek-moe-16b-chat, then neurobrix serve --model deepseek-moe-16b-chat.
CLI Usage
$ neurobrix import deepseek-ai/deepseek-moe-16b-chat
$ neurobrix serve --model deepseek-moe-16b-chat
$ neurobrix run --prompt "your prompt here"
$ neurobrix stop