RunsOnMy ModelsGPUs & MacsAPI pricesVRAM calculatorNew

Can I run Nemotron 3.5 Parse locally?

Nemotron 3.5 Parse by nvidia has 631M parameters. At 4-bit with an 8K context it needs about 1.1 GB of GPU memory. It fits on 65 of 65 GPUs and Macs we track.

1.1 GBQ4_K_M · 8K context
RTX 4060smallest consumer GPU that fits
M1 16GBsmallest Mac that fits
16Kmax context
GQAattention type

Memory needed by quant and context length

GB of GPU memory = weights + KV cache (FP16) + runtime overhead. Weights use real file sizes from SkyContact/Nemotron-3.5-Parse-GGUF where available (marked •), otherwise parameter count × bits per weight. How we calculate.

QuantQuality4K ctx8K ctx
2-bit (Q2_K)rough0.8 GB0.9 GB
3-bit (Q3_K_M)usable0.9 GB1.0 GB
4-bit (Q4_K_M)good, most popular1.0 GB1.1 GB
5-bit (Q5_K_M)very good1.0 GB1.1 GB
6-bit (Q6_K)near lossless1.1 GB1.2 GB
8-bit (Q8_0) •effectively lossless1.4 GB1.5 GB
16-bit (F16/BF16) •original weights1.8 GB1.9 GB

Which GPUs and Macs run Nemotron 3.5 Parse?

Best quality quant that fits at 8K context. "Tight" = within 10% of usable memory. Speed is a ceiling from memory bandwidth (real speed is typically 50–80% of it).

HardwareMemoryFits?Best quantNeedsMax tok/s (est.)
RTX 3060 12GB12 GBFitsF161.9 GB275details
RTX 4060 Ti 16GB16 GBFitsF161.9 GB220details
RTX 309024 GBFitsF161.9 GB716details
RTX 409024 GBFitsF161.9 GB771details
RTX 509032 GBFitsF161.9 GB1371details
RTX 5070 Ti16 GBFitsF161.9 GB686details
RX 7900 XTX24 GBFitsF161.9 GB735details
Mac M4 Pro 48GB48 GBFitsF161.9 GB209details
Mac M4 Max 128GB128 GBFitsF161.9 GB418details
Mac M3 Ultra 512GB512 GBFitsF161.9 GB627details
DGX Spark128 GBFitsF161.9 GB209details
RTX 407012 GBFitsF161.9 GB386details
RTX 5060 Ti 16GB16 GBFitsF161.9 GB343details
RTX 40608 GBFitsF161.9 GB208details
Arc B58012 GBFitsF161.9 GB349details
RTX 3060 Ti8 GBFitsF161.9 GB343details
RTX 30708 GBFitsF161.9 GB343details
RTX 3080 10GB10 GBFitsF161.9 GB581details
RTX 3080 12GB12 GBFitsF161.9 GB698details
RTX 3090 Ti24 GBFitsF161.9 GB771details
RTX 4060 Ti 8GB8 GBFitsF161.9 GB220details
RTX 4070 Super12 GBFitsF161.9 GB386details
RTX 4070 Ti12 GBFitsF161.9 GB386details
RTX 4070 Ti Super16 GBFitsF161.9 GB514details
RTX 408016 GBFitsF161.9 GB549details
RTX 4080 Super16 GBFitsF161.9 GB563details
RTX 50608 GBFitsF161.9 GB343details
RTX 507012 GBFitsF161.9 GB514details
RTX 508016 GBFitsF161.9 GB735details
2x RTX 30902×24 GBFitsF161.9 GB716details
2x RTX 40902×24 GBFitsF161.9 GB771details
2x RTX 50902×32 GBFitsF161.9 GB1371details
RTX A600048 GBFitsF161.9 GB588details
RTX 6000 Ada48 GBFitsF161.9 GB735details
RTX PRO 6000 Blackwell96 GBFitsF161.9 GB1371details
L40S48 GBFitsF161.9 GB661details
A100 40GB40 GBFitsF161.9 GB1190details
A100 80GB80 GBFitsF161.9 GB1560details
H100 PCIe80 GBFitsF161.9 GB1530details
H100 SXM80 GBFitsF161.9 GB2563details
H200141 GBFitsF161.9 GB3673details
RX 76008 GBFitsF161.9 GB220details
RX 7700 XT12 GBFitsF161.9 GB331details
RX 7800 XT16 GBFitsF161.9 GB477details
RX 7900 GRE16 GBFitsF161.9 GB441details
RX 7900 XT20 GBFitsF161.9 GB612details
RX 907016 GBFitsF161.9 GB490details
RX 9070 XT16 GBFitsF161.9 GB490details
Ryzen AI Max+ 395 128GB128 GBFitsF161.9 GB196details
Arc A770 16GB16 GBFitsF161.9 GB428details
Mac M1 16GB16 GBFitsF161.9 GB52details
Mac M2 24GB24 GBFitsF161.9 GB77details
Mac M3 24GB24 GBFitsF161.9 GB77details
Mac M4 16GB16 GBFitsF161.9 GB92details
Mac M4 32GB32 GBFitsF161.9 GB92details
Mac M1 Pro 32GB32 GBFitsF161.9 GB153details
Mac M3 Pro 36GB36 GBFitsF161.9 GB115details
Mac M4 Pro 24GB24 GBFitsF161.9 GB209details
Mac M4 Pro 64GB64 GBFitsF161.9 GB209details
Mac M1 Max 64GB64 GBFitsF161.9 GB306details
Mac M2 Max 96GB96 GBFitsF161.9 GB306details
Mac M3 Max 128GB128 GBFitsF161.9 GB306details
Mac M4 Max 64GB64 GBFitsF161.9 GB418details
Mac M2 Ultra 192GB192 GBFitsF161.9 GB612details
Mac M3 Ultra 256GB256 GBFitsF161.9 GB627details

Questions

How much VRAM does Nemotron 3.5 Parse need?

About 1.1 GB at 4-bit (Q4_K_M) with an 8K context: 0.4 GB of weights (estimated), 0.2 GB of KV cache and 0.5 GB of runtime overhead. At 8-bit it needs about 1.5 GB.

What is the smallest GPU that runs Nemotron 3.5 Parse?

At Q4_K_M and 8K context, the smallest single consumer GPU in our table that fits it is the RTX 4060 (8 GB).

Can I run Nemotron 3.5 Parse on a Mac?

Yes. At Q4_K_M with 8K context it fits on a Mac M1 16GB or larger, using the default macOS GPU memory limit.

Model facts

Hugging Face reponvidia/Nemotron-3.5-Parse
Parameters630,800,504 (631M)
Layers / hidden size10 / 1024
AttentionGQA; 16 heads, 8 KV heads, head dim 64
KV cache per token (FP16)20.0 KiB
Max context16,384 tokens
Licenseopenmdw-1.1
Released on HFSep 3, 2026
Downloads (30d) / likes25,540 / 169

Sources: Hugging Face model API · config.json (nvidia/Nemotron-3.5-Parse) · GGUF files · GPU specs from vendor pages (list). Data fetched Nov 17, 2026. Spot an error? Report it.