Skip to content

Seed OSS 36B Instruct VRAM requirements

How much VRAM Seed OSS 36B Instruct needs
4-bit weights: 20.4 GBWith 8K context: 23.3 GBWith 32K context: 29.3 GB8-bit, 8K context: 38.7 GB
Parameters
36.2B
Longest context
512K
Released
Aug 2025
License
Apache 2.0
Strongest at
Not rated

How much VRAM Seed OSS 36B Instruct needs

Memory Seed OSS 36B Instruct needs at each precision and context length
PrecisionWeightsWith 8KWith 32KWith 128K
16-bit67.3 GB70.2 GB76.2 GB100 GB
8-bit35.8 GB38.7 GB44.7 GB68.7 GB
6-bit27.6 GB30.5 GB36.5 GB60.5 GB
5-bit23.9 GB26.8 GB32.8 GB56.8 GB
4-bit20.4 GB23.3 GB29.3 GB53.3 GB
3-bit16.5 GB19.4 GB25.4 GB49.4 GB
2-bit12.6 GB15.5 GB21.5 GB45.5 GB
1.6-bit6.9 GB9.8 GB15.8 GB39.8 GB

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

GPUs and Macs that run Seed OSS 36B Instruct

The best way Seed OSS 36B Instruct runs on common GPUs, Macs and AI boxes
MachineBest fitContextSpeedFirst replyReport
RTX 4060Too bigPDF: Seed OSS 36B Instruct on RTX 4060 report
RTX 5060Too bigPDF: Seed OSS 36B Instruct on RTX 5060 report
RTX 3070Too bigPDF: Seed OSS 36B Instruct on RTX 3070 report
RTX 3080 10 GBToo bigPDF: Seed OSS 36B Instruct on RTX 3080 10 GB report
RTX 3060 12 GBToo bigPDF: Seed OSS 36B Instruct on RTX 3060 12 GB report
Arc B580Too bigPDF: Seed OSS 36B Instruct on Arc B580 report
RTX 4070Too bigPDF: Seed OSS 36B Instruct on RTX 4070 report
RTX 4070 SuperToo bigPDF: Seed OSS 36B Instruct on RTX 4070 Super report
RTX 5070Too bigPDF: Seed OSS 36B Instruct on RTX 5070 report
RTX 4060 Ti 16 GBToo bigPDF: Seed OSS 36B Instruct on RTX 4060 Ti 16 GB report
RTX 5060 Ti 16 GBToo bigPDF: Seed OSS 36B Instruct on RTX 5060 Ti 16 GB report
Arc A770 16 GBToo bigPDF: Seed OSS 36B Instruct on Arc A770 16 GB report
Radeon RX 9070 XTToo bigPDF: Seed OSS 36B Instruct on Radeon RX 9070 XT report
RTX 4070 Ti SuperToo bigPDF: Seed OSS 36B Instruct on RTX 4070 Ti Super report
RTX 4080 SuperToo bigPDF: Seed OSS 36B Instruct on RTX 4080 Super report
RTX 5070 TiToo bigPDF: Seed OSS 36B Instruct on RTX 5070 Ti report
RTX 5080Too bigPDF: Seed OSS 36B Instruct on RTX 5080 report
Radeon RX 7900 XTX3-bit20K30 tok/s3.4 sPDF: Seed OSS 36B Instruct on Radeon RX 7900 XTX report
RTX 30903-bit20K35 tok/s2.1 sPDF: Seed OSS 36B Instruct on RTX 3090 report
RTX 40903-bit20K38 tok/s0.9 sPDF: Seed OSS 36B Instruct on RTX 4090 report
Radeon AI PRO R97005-bit20K15 tok/s2.2 sPDF: Seed OSS 36B Instruct on Radeon AI PRO R9700 report
RTX 50905-bit20K46 tok/s0.7 sPDF: Seed OSS 36B Instruct on RTX 5090 report
2× RTX 30908-bit32K17 tok/s1.2 sPDF: Seed OSS 36B Instruct on 2× RTX 3090 report
2× RTX 40908-bit32K18 tok/s0.5 sPDF: Seed OSS 36B Instruct on 2× RTX 4090 report
RTX PRO 5000 Blackwell 48 GB8-bit32K24 tok/s1.1 sPDF: Seed OSS 36B Instruct on RTX PRO 5000 Blackwell 48 GB report
RTX 6000 Ada8-bit32K17 tok/s0.8 sPDF: Seed OSS 36B Instruct on RTX 6000 Ada report
L40S8-bit32K16 tok/s1.3 sPDF: Seed OSS 36B Instruct on L40S report
2× RTX 50908-bit88K31 tok/s0.4 sPDF: Seed OSS 36B Instruct on 2× RTX 5090 report
4× RTX 30908-bit208K17 tok/s0.7 sPDF: Seed OSS 36B Instruct on 4× RTX 3090 report
RTX PRO 6000 Blackwell8-bit216K31 tok/s0.6 sPDF: Seed OSS 36B Instruct on RTX PRO 6000 Blackwell report
Mac mini M4 16 GBToo bigPDF: Seed OSS 36B Instruct on Mac mini M4 16 GB report
Mac mini M4 24 GBToo bigPDF: Seed OSS 36B Instruct on Mac mini M4 24 GB report
Mac mini M4 32 GBToo bigPDF: Seed OSS 36B Instruct on Mac mini M4 32 GB report
Mac mini M4 Pro 24 GBToo bigPDF: Seed OSS 36B Instruct on Mac mini M4 Pro 24 GB report
Mac mini M4 Pro 48 GB6-bit20K7.4 tok/s24 sPDF: Seed OSS 36B Instruct on Mac mini M4 Pro 48 GB report
Mac mini M4 Pro 64 GB8-bit36K5.8 tok/s24 sPDF: Seed OSS 36B Instruct on Mac mini M4 Pro 64 GB report
M4 Max 128 GB8-bit216K10 tok/s12 sPDF: Seed OSS 36B Instruct on M4 Max 128 GB report
M5 Max 128 GB8-bit216K12 tok/s3.5 sPDF: Seed OSS 36B Instruct on M5 Max 128 GB report
M3 Ultra 96 GB8-bit120K13 tok/s6.8 sPDF: Seed OSS 36B Instruct on M3 Ultra 96 GB report
M3 Ultra 256 GB8-bit512K13 tok/s6.8 sPDF: Seed OSS 36B Instruct on M3 Ultra 256 GB report
M3 Ultra 512 GB8-bit512K13 tok/s6.8 sPDF: Seed OSS 36B Instruct on M3 Ultra 512 GB report
M5 Ultra 512 GB8-bit512K19 tok/s2.1 sPDF: Seed OSS 36B Instruct on M5 Ultra 512 GB report
Ryzen AI Max+ 395 64 GB8-bit36K4.0 tok/s6.9 sPDF: Seed OSS 36B Instruct on Ryzen AI Max+ 395 64 GB report
Ryzen AI Max+ 395 128 GB8-bit216K4.0 tok/s6.9 sPDF: Seed OSS 36B Instruct on Ryzen AI Max+ 395 128 GB report
RTX Spark8-bit288K5.5 tok/s2.4 sPDF: Seed OSS 36B Instruct on RTX Spark report
DGX Spark8-bit288K5.0 tok/s2.4 sPDF: Seed OSS 36B Instruct on DGX Spark report
Dual DGX Spark8-bit512K5.0 tok/s1.4 sPDF: Seed OSS 36B Instruct on Dual DGX Spark report
DGX Station GB3008-bit512K105 tok/s0.2 sPDF: Seed OSS 36B Instruct on DGX Station GB300 report
2× RTX PRO 6000 Blackwell8-bit512K31 tok/s0.3 sPDF: Seed OSS 36B Instruct on 2× RTX PRO 6000 Blackwell report
4× RTX PRO 6000 Blackwell8-bit512K31 tok/s0.2 sPDF: Seed OSS 36B Instruct on 4× RTX PRO 6000 Blackwell report
8× RTX PRO 6000 Blackwell8-bit512K31 tok/s0.1 sPDF: Seed OSS 36B Instruct on 8× RTX PRO 6000 Blackwell report
A100 80 GB SXM8-bit152K35 tok/s1.5 sPDF: Seed OSS 36B Instruct on A100 80 GB SXM report
H100 SXM8-bit152K56 tok/s0.5 sPDF: Seed OSS 36B Instruct on H100 SXM report
4× H200 NVL8-bit512K76 tok/s0.2 sPDF: Seed OSS 36B Instruct on 4× H200 NVL report
8× H200 NVL8-bit512K76 tok/s0.1 sPDF: Seed OSS 36B Instruct 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 unsloth/Seed-OSS-36B-Instruct-GGUF
vLLM
vllm serve ByteDance-Seed/Seed-OSS-36B-Instruct
MLX (Mac)
mlx_lm.chat --model mlx-community/Seed-OSS-36B-Instruct-4bit

Questions

How much VRAM does Seed OSS 36B Instruct need?

Seed OSS 36B Instruct needs 20.4 GB for its weights at 4-bit, 23.3 GB with an 8K context and 29.3 GB with 32K. At 8-bit, 38.7 GB.

What is the minimum VRAM for Seed OSS 36B Instruct?

About 19.4 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 Radeon RX 7900 XTX.

How big is Seed OSS 36B Instruct?

36.2B parameters: 67.3 GB of weights at 16-bit, 20.4 GB at 4-bit.

Can Seed OSS 36B Instruct run on an RTX 4090?

Yes. At 3-bit it holds 20K of context and writes about 38 tokens a second.

More models