Skip to content

OneRec 1.7B VRAM requirements

How much VRAM OneRec 1.7B needs
4-bit weights: 1.2 GBWith 8K context: 2.9 GBWith 32K context: 5.6 GB8-bit, 8K context: 3.8 GB
Parameters
2.1B
Longest context
40K
Released
Dec 2025
License
Apache 2.0
Strongest at
Not rated

How much VRAM OneRec 1.7B needs

Memory OneRec 1.7B needs at each precision and context length
PrecisionWeightsWith 8KWith 32KWith 40K
16-bit4.0 GB5.7 GB8.3 GB9.2 GB
8-bit2.1 GB3.8 GB6.5 GB7.3 GB
6-bit1.6 GB3.4 GB6.0 GB6.9 GB
5-bit1.4 GB3.1 GB5.8 GB6.6 GB
4-bit1.2 GB2.9 GB5.6 GB6.4 GB
3-bit1.0 GB2.7 GB5.3 GB6.2 GB
2-bit762 MB2.5 GB5.1 GB6.0 GB
1.6-bit419 MB2.1 GB4.8 GB5.6 GB

One conversation, 16-bit context cache, runtime overhead included.

GPUs and Macs that run OneRec 1.7B

The best way OneRec 1.7B runs on common GPUs, Macs and AI boxes
MachineBest fitContextSpeedFirst replyReport
RTX 40608-bit40K79 tok/s0.3 sPDF: OneRec 1.7B on RTX 4060 report
RTX 50608-bit40K124 tok/s0.2 sPDF: OneRec 1.7B on RTX 5060 report
RTX 30708-bit40K124 tok/s0.2 sPDF: OneRec 1.7B on RTX 3070 report
RTX 3080 10 GB8-bit40K196 tok/s0.1 sPDF: OneRec 1.7B on RTX 3080 10 GB report
RTX 3060 12 GB8-bit40K102 tok/s0.3 sPDF: OneRec 1.7B on RTX 3060 12 GB report
Arc B5808-bit40K86 tok/s0.2 sPDF: OneRec 1.7B on Arc B580 report
RTX 40708-bit40K138 tok/s0.1 sPDF: OneRec 1.7B on RTX 4070 report
RTX 4070 Super8-bit40K138 tok/s0.1 sPDF: OneRec 1.7B on RTX 4070 Super report
RTX 50708-bit40K177 tok/s0.1 sPDF: OneRec 1.7B on RTX 5070 report
RTX 4060 Ti 16 GB8-bit40K83 tok/s0.2 sPDF: OneRec 1.7B on RTX 4060 Ti 16 GB report
RTX 5060 Ti 16 GB8-bit40K124 tok/s0.2 sPDF: OneRec 1.7B on RTX 5060 Ti 16 GB report
Arc A770 16 GB8-bit40K102 tok/s0.2 sPDF: OneRec 1.7B on Arc A770 16 GB report
Radeon RX 9070 XT8-bit40K137 tok/s0.1 sPDF: OneRec 1.7B on Radeon RX 9070 XT report
RTX 4070 Ti Super8-bit40K177 tok/s0.1 sPDF: OneRec 1.7B on RTX 4070 Ti Super report
RTX 4080 Super8-bit40K191 tok/s0.1 sPDF: OneRec 1.7B on RTX 4080 Super report
RTX 5070 Ti8-bit40K225 tok/s0.1 sPDF: OneRec 1.7B on RTX 5070 Ti report
RTX 50808-bit40K237 tok/s0.1 sPDF: OneRec 1.7B on RTX 5080 report
Radeon RX 7900 XTX8-bit40K188 tok/s0.2 sPDF: OneRec 1.7B on Radeon RX 7900 XTX report
RTX 30908-bit40K233 tok/s0.1 sPDF: OneRec 1.7B on RTX 3090 report
RTX 40908-bit40K247 tok/s0.1 sPDF: OneRec 1.7B on RTX 4090 report
Radeon AI PRO R97008-bit40K137 tok/s0.1 sPDF: OneRec 1.7B on Radeon AI PRO R9700 report
RTX 50908-bit40K378 tok/s0.1 sPDF: OneRec 1.7B on RTX 5090 report
2× RTX 30908-bit40K233 tok/s0.1 sPDF: OneRec 1.7B on 2× RTX 3090 report
2× RTX 40908-bit40K247 tok/s0.1 sPDF: OneRec 1.7B on 2× RTX 4090 report
RTX PRO 5000 Blackwell 48 GB8-bit40K308 tok/s0.1 sPDF: OneRec 1.7B on RTX PRO 5000 Blackwell 48 GB report
RTX 6000 Ada8-bit40K237 tok/s0.1 sPDF: OneRec 1.7B on RTX 6000 Ada report
L40S8-bit40K218 tok/s0.1 sPDF: OneRec 1.7B on L40S report
2× RTX 50908-bit40K378 tok/s0.1 sPDF: OneRec 1.7B on 2× RTX 5090 report
4× RTX 30908-bit40K233 tok/s0.1 sPDF: OneRec 1.7B on 4× RTX 3090 report
RTX PRO 6000 Blackwell8-bit40K378 tok/s0.1 sPDF: OneRec 1.7B on RTX PRO 6000 Blackwell report
Mac mini M4 16 GB8-bit40K40 tok/s2.6 sPDF: OneRec 1.7B on Mac mini M4 16 GB report
Mac mini M4 24 GB8-bit40K40 tok/s2.6 sPDF: OneRec 1.7B on Mac mini M4 24 GB report
Mac mini M4 32 GB8-bit40K40 tok/s2.6 sPDF: OneRec 1.7B on Mac mini M4 32 GB report
Mac mini M4 Pro 24 GB8-bit40K81 tok/s1.3 sPDF: OneRec 1.7B on Mac mini M4 Pro 24 GB report
Mac mini M4 Pro 48 GB8-bit40K81 tok/s1.3 sPDF: OneRec 1.7B on Mac mini M4 Pro 48 GB report
Mac mini M4 Pro 64 GB8-bit40K81 tok/s1.3 sPDF: OneRec 1.7B on Mac mini M4 Pro 64 GB report
M4 Max 128 GB8-bit40K129 tok/s0.6 sPDF: OneRec 1.7B on M4 Max 128 GB report
M5 Max 128 GB8-bit40K140 tok/s0.2 sPDF: OneRec 1.7B on M5 Max 128 GB report
M3 Ultra 96 GB8-bit40K154 tok/s0.4 sPDF: OneRec 1.7B on M3 Ultra 96 GB report
M3 Ultra 256 GB8-bit40K154 tok/s0.4 sPDF: OneRec 1.7B on M3 Ultra 256 GB report
M3 Ultra 512 GB8-bit40K154 tok/s0.4 sPDF: OneRec 1.7B on M3 Ultra 512 GB report
M5 Ultra 512 GB8-bit40K194 tok/s0.1 sPDF: OneRec 1.7B on M5 Ultra 512 GB report
Ryzen AI Max+ 395 64 GB8-bit40K62 tok/s0.4 sPDF: OneRec 1.7B on Ryzen AI Max+ 395 64 GB report
Ryzen AI Max+ 395 128 GB8-bit40K62 tok/s0.4 sPDF: OneRec 1.7B on Ryzen AI Max+ 395 128 GB report
RTX Spark8-bit40K86 tok/s0.1 sPDF: OneRec 1.7B on RTX Spark report
DGX Spark8-bit40K79 tok/s0.1 sPDF: OneRec 1.7B on DGX Spark report
Dual DGX Spark8-bit40K79 tok/s0.1 sPDF: OneRec 1.7B on Dual DGX Spark report
DGX Station GB3008-bit40K772 tok/s0.1 sPDF: OneRec 1.7B on DGX Station GB300 report
2× RTX PRO 6000 Blackwell8-bit40K378 tok/s0.1 sPDF: OneRec 1.7B on 2× RTX PRO 6000 Blackwell report
4× RTX PRO 6000 Blackwell8-bit40K378 tok/s0.1 sPDF: OneRec 1.7B on 4× RTX PRO 6000 Blackwell report
8× RTX PRO 6000 Blackwell8-bit40K378 tok/s0.1 sPDF: OneRec 1.7B on 8× RTX PRO 6000 Blackwell report
A100 80 GB SXM8-bit40K412 tok/s0.1 sPDF: OneRec 1.7B on A100 80 GB SXM report
H100 SXM8-bit40K554 tok/s0.1 sPDF: OneRec 1.7B on H100 SXM report
4× H200 NVL8-bit40K660 tok/s0.1 sPDF: OneRec 1.7B on 4× H200 NVL report
8× H200 NVL8-bit40K660 tok/s0.1 sPDF: OneRec 1.7B on 8× H200 NVL report

Best fit is the highest quality, up to 8-bit, that leaves 16K of context. First reply is the wait for a 2K prompt. Speeds are rough, plus or minus 25%.

How to run it

llama.cpp
llama-server -hf mradermacher/OneRec-1.7B-i1-GGUF
vLLM
vllm serve OpenOneRec/OneRec-1.7B

Questions

How much VRAM does OneRec 1.7B need?

OneRec 1.7B needs 1.2 GB for its weights at 4-bit, 2.9 GB with an 8K context and 5.6 GB with 32K. At 8-bit, 3.8 GB.

What is the minimum VRAM for OneRec 1.7B?

About 2.7 GB at 3-bit with an 8K context, the lowest precision we recommend for a model this size. The smallest machine here that runs it without spilling is the RTX 4060.

How big is OneRec 1.7B?

2.1B parameters: 4.0 GB of weights at 16-bit, 1.2 GB at 4-bit.

Can OneRec 1.7B run on an RTX 4090?

Yes. At 8-bit it holds 40K of context and writes about 247 tokens a second.

More models