Nemotron 3.5 Rerank 8B

About 6.1 GB at Q4, 10.3 GB at Q8, and 18.9 GB at FP16. All figures are estimates from the 8B parameter count, covering weights plus about 15% and 0.5 GB for a short context.

Yes on 24 GB, 16 GB and 12 GB GPUs, and also on 8 GB at lower precision. Estimates: 24 GB fits at FP16 (18.9 GB), 16 GB and 12 GB fit at Q8 (10.3 GB), and 8 GB fits at Q4 (6.1 GB).

Yes. The estimated VRAM at Q4 is 6.1 GB. Assuming macOS lets the GPU use about 75% of unified memory, the smallest common Mac memory size that fits Q4 is 16 GB.

Yes, you can self-host Nemotron 3.5 Rerank 8B for full control of your data, as the model has open weights.

Run Nemotron 3.5 Rerank 8B locally

The context window is 33K tokens (32,768), and the maximum output is 29K tokens (29,491).

The closest alternatives, matched on model kind and release date with no comparable price, are Qwen3 Reranker 0.6B by Qwen, Rerank 5 Pro by Cohere and Rerank 5 Fast by Cohere.

Qwen3 Reranker 0.6BNemotron 3.5 Rerank 8B vs Qwen3 Reranker 0.6BRerank 5 ProNemotron 3.5 Rerank 8B vs Rerank 5 ProRerank 5 FastNemotron 3.5 Rerank 8B vs Rerank 5 Fast

Nemotron 3.5 Rerank 8B has a 33K token (32,768) context, costs $0.00 input and $0.00 output per 1M tokens, and was released October 21, 2026. Nemotron 3.5 ASR Streaming Multilingual 0.6B costs $3.33 input and $0.00 output per 1M tokens and was released August 13, 2026.

Nemotron 3.5 ASR Streaming Multilingual 0.6BNemotron 3.5 Rerank 8B vs Nemotron 3.5 ASR Streaming Multilingual 0.6B

Nemotron 3.5 Rerank 8B is open weights, not open source.

You can download the weights from the Hugging Face repository nvidia/Nemotron-3.5-Rerank-8B-BF16.

Run Nemotron 3.5 Rerank 8B locally

NVIDIA made Nemotron 3.5 Rerank 8B, and it was first listed on OpenRouter on October 21, 2026.

More from NVIDIA

Nemotron 3.5 Rerank 8B has open weights.