nemotron-cascade-2:30b-a3b-mxfp8

161.4K 1 week ago

An open 30B MoE model from NVIDIA with 3B activated parameters that delivers strong reasoning and agentic capabilities.

tools thinking 30b
ollama run nemotron-cascade-2:30b-a3b-mxfp8

Details

1 week ago

7a05ba2133bd ยท 33GB

{ "architectures": [ "NemotronHForCausalLM" ], "attention_bias": false, "attention_dropout": 0.0, "a
{ "_from_model_config": true, "bos_token_id": 1, "eos_token_id": [2, 11], "pad_token_id": 0, "transf
{ "producer": { "name": "modelopt", "version": "0.0.0" }, "quantization": { "quant_algo": "MXFP8", "
{ "bos_token": { "content": "<s>", "lstrip": false, "normalized": false, "rstrip": false, "single_wo
{ "version": "1.0", "truncation": null, "padding": null, "added_tokens": [ { "id": 0, "content": "<u
{ "add_bos_token": false, "add_eos_token": false, "add_prefix_space": false, "added_tokens_decoder":
NVIDIA Open Model License Agreement Last Modified: October 24, 2025 This NVIDIA Open Model License A
{ "temperature": 1, "top_p": 0.95 }
{{ .Prompt }}
6243 tensors

Readme

Nemotron-Cascade-2-30B-A3B is an open 30B MoE model with 3B activated parameters that delivers strong reasoning and agentic capabilities. It achieves gold medal performance in both the 2025 International Mathematical Olympiad (IMO) and the International Olympiad in Informatics (IOI). It operates in both thinking and instruct (non-reasoning) modes.

image.png