← CookbookView source on GitHub ↗

Models on Nebius Token Factory

Browse and run the latest open models at tokenfactory.nebius.com/models.


Click on the model name for model guide, and ▶ to Try it @ TF in the playground.

View all available NVIDIA Nemotron Models

Model Provider Parameters Context Highlights
Kimi-K3

▶ Try it
Moonshot AI 2.8 T 1 M Latest open model from Moonshot AI — strong reasoning and agentic capability
Kimi K2.7 Code

▶ Try it
Moonshot AI 1T total · 32B active 256K Coding-focused agentic MoE — native multimodal with MoonViT; ~30% thinking-token reduction vs K2.6
MiniMax M3

▶ Try it
MiniMax 428B total · 23B active 1 M Native multimodal MoE (text+image+video) with MiniMax Sparse Attention — frontier coding and cowork
DeepSeek V4 Pro

▶ Try it
DeepSeek 1.6T total · 49B active 1M Frontier open model from DeepSeek
GLM-5.2

▶ Try it
Z.ai 753 B 1 M improved reasoning, agentic capability, and tool use
Qwen3.5-397B-A17B

▶ Try it
Alibaba / Qwen 397B total · 17B active 262K Latest Qwen MoE — best-in-class reasoning and coding
Nemotron-3-Super-120B-A12B

▶ Try it
NVIDIA 120B total · 12B active 128K Efficient MoE optimized for enterprise inference