LocalFit
Submit a benchmark

Nemotron-Labs-Diffusion-8B

nvidia/Nemotron-Labs-Diffusion-8B

ReleasedRelease date unknownSize8.5B tensors · denseRoleAssistantLicensenvidia-nemotron-open-model-licenseContext256K config max

Where it runs 8K context · ≥ 5 t/s floor

SetupBest planMemoryGenerate
8 GB GPU + 32 GB RAMQ5_K_Mllama.cpp CUDAFull GPUKV unknown~27–39 t/s Estimated
12 GB GPU + 32 GB RAMQ5_K_Mllama.cpp CUDAFull GPUKV unknown~45–49 t/s Estimated
16 GB GPU + 64 GB RAMQ5_K_Mllama.cpp CUDAFull GPUKV unknown~94–115 t/s Estimated
24 GB GPU + 64 GB RAMQ5_K_Mllama.cpp CUDAFull GPUKV unknown~95–119 t/s Estimated
32 GB GPU + 128 GB RAMQ5_K_Mllama.cpp CUDAFull GPUKV unknown~149–238 t/s Estimated
2× 24 GB GPUs + 128 GB RAMQ5_K_Mllama.cpp CUDAFull GPUKV unknown~87–112 t/s Estimated
96 GB workstation GPU + 256 GB RAMQ5_K_Mllama.cpp CUDAFull GPUKV unknown~153–240 t/s Estimated
Mac, 24 GBQ5_K_Mllama.cpp MetalUnified memoryKV unknown~32–38 t/s Estimated
Mac, 48 GBQ5_K_Mllama.cpp MetalUnified memoryKV unknown~32–38 t/s Estimated
Mac, 128 GBQ5_K_Mllama.cpp MetalUnified memoryKV unknown~52–79 t/s Estimated
Mac, 512 GBQ5_K_Mllama.cpp MetalUnified memoryKV unknown~58–89 t/s Estimated
DGX Spark 128 GBQ5_K_Mllama.cpp CUDAUnified memoryKV unknown~34–38 t/s Estimated
Ryzen AI Max+ 395, 128 GBQ5_K_Mllama.cpp VulkanUnified memoryKV unknown~23–34 t/s Estimated
CPU only, 64 GB DDR5-5600 dual channelQ4_K_Mllama.cpp CPUCPU onlyKV unknown~3.9–4.6 t/s Estimated extrapolated

Same plan engine as the planner, with its default reserves and RAM. How plans are chosen.

Files exact Hugging Face bytes

ArtifactSizeFilesRevisionSource
Q5_K_M
suzukimain/Nemotron-Labs-Diffusion-8B-GGUF
5.64 GiB1ee1f33abd1HF files
Q4_K_M
suzukimain/Nemotron-Labs-Diffusion-8B-GGUF
4.84 GiB1ee1f33abd1HF files
Native safetensors BF16
nvidia/Nemotron-Labs-Diffusion-8B
15.81 GiB116c67f0560HF files

Not found in the representative quant search (not proven unavailable): GGUF FP16, GGUF Q8_0, MLX 8-bit, MLX 4-bit, EXL3, AWQ, GPTQ. Complete selected file sets; projectors, MTP/draft and lookup files are listed separately and budgeted only by plans that use them. File bytes are not allocated memory.

Quality

No Artificial Analysis evaluation mapped to this checkpoint.

Arena text rating

Unknown

Publisher-reported benchmarks

None extracted from the model card.

AA and Arena describe the hosted model in the stated mode, not each quantized file; scores are never inherited by base, pruned, abliterated or quantized variants and never averaged together.

How to run it

Transformers custommodeling · unknown

publisher-documented, not executed source

from transformers import AutoModel, AutoTokenizer
import torch

repo_name = "nvidia/Nemotron-Labs-Diffusion-8B"

tokenizer = AutoTokenizer.from_pretrained(repo_name, trust_remote_code=True)
model = AutoModel.from_pretrained(repo_name, trust_remote_code=True)
model = model.cuda().to(torch.bfloat16)

history = []

user_input = input("User: ").strip()
history.append({"role": "user", "content": user_input})

prompt = tokenizer.apply_chat_template(history, tokenize=False, add_generation_prompt=True)
prompt_ids = tokenizer(prompt, return_tensors='pt').input_ids.to(device='cuda')

## Chat in AR Mode
out_ids, nfe = model.ar_generate(inputs.input_ids, max_new_tokens=512)

## Chat in dLM Mode
out_ids, nfe = model.generate(prompt_ids, max_new_tokens=512, block_length=32, threshold=0.9, eos_token_id=tokenizer.eos_token_id)

## Chat in Linear Self-Speculation Mode
out_ids, nfe = model.linear_spec_generate(prompt_ids, max_new_tokens=512, block_length=32, eos_token_id=tokenizer.eos_token_id)

tokenized_out = tokenizer.batch_decode(out_ids[:, prompt_ids.shape[1]:], skip_special_tokens=True)[0]
print(f"Model: {tokenized_out}")
print(f"[Num Function Eval (NFE)={nfe}]")

Prerequisites: transformers>=5.0.0; custom code as shown in card.

  • Tri-mode autoregressive/diffusion/block diffusion generation; compare speed only with same mode, steps, block size.
  • Custom remote-code cache/denoising state is not generic KV.

Recipes are source examples (settings are the publisher’s or author’s, not measured optima) and were not executed by LocalFit. A downloadable file does not prove your backend build supports the architecture.

Reported speeds 0

No reported measurement yet. Submit one.