Skip to content

Academic Ds 9B VRAM requirements

How much VRAM Academic Ds 9B needs
4-bit weights: 5.3 GBWith 8K context: 6.2 GB8-bit, 8K context: 10.2 GB
Parameters
9.4B
Active per token
1.6B
Longest context
8K
Released
Apr 2025
License
Apache 2.0
Strongest at
Not rated

How much VRAM Academic Ds 9B needs

Memory Academic Ds 9B needs at each precision and context length
PrecisionWeightsWith 8K
16-bit17.4 GB18.3 GB
8-bit9.3 GB10.2 GB
6-bit7.2 GB8.1 GB
5-bit6.2 GB7.1 GB
4-bit5.3 GB6.2 GB
3-bit4.3 GB5.2 GB
2-bit3.3 GB4.2 GB
1.6-bit1.8 GB2.7 GB

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

GPUs and Macs that run Academic Ds 9B

The best way Academic Ds 9B runs on common GPUs, Macs and AI boxes
MachineBest fitContextSpeedFirst replyReport
RTX 40605-bit8K143 tok/s0.5 sPDF: Academic Ds 9B on RTX 4060 report
RTX 50605-bit8K216 tok/s0.4 sPDF: Academic Ds 9B on RTX 5060 report
RTX 30705-bit8K216 tok/s0.4 sPDF: Academic Ds 9B on RTX 3070 report
RTX 3080 10 GB6-bit8K290 tok/s0.3 sPDF: Academic Ds 9B on RTX 3080 10 GB report
RTX 3060 12 GB8-bit8K129 tok/s0.6 sPDF: Academic Ds 9B on RTX 3060 12 GB report
Arc B5808-bit8K110 tok/s0.4 sPDF: Academic Ds 9B on Arc B580 report
RTX 40708-bit8K172 tok/s0.3 sPDF: Academic Ds 9B on RTX 4070 report
RTX 4070 Super8-bit8K172 tok/s0.2 sPDF: Academic Ds 9B on RTX 4070 Super report
RTX 50708-bit8K217 tok/s0.3 sPDF: Academic Ds 9B on RTX 5070 report
RTX 4060 Ti 16 GB8-bit8K106 tok/s0.4 sPDF: Academic Ds 9B on RTX 4060 Ti 16 GB report
RTX 5060 Ti 16 GB8-bit8K155 tok/s0.3 sPDF: Academic Ds 9B on RTX 5060 Ti 16 GB report
Arc A770 16 GB8-bit8K129 tok/s0.4 sPDF: Academic Ds 9B on Arc A770 16 GB report
Radeon RX 9070 XT8-bit8K169 tok/s0.2 sPDF: Academic Ds 9B on Radeon RX 9070 XT report
RTX 4070 Ti Super8-bit8K217 tok/s0.2 sPDF: Academic Ds 9B on RTX 4070 Ti Super report
RTX 4080 Super8-bit8K233 tok/s0.2 sPDF: Academic Ds 9B on RTX 4080 Super report
RTX 5070 Ti8-bit8K271 tok/s0.2 sPDF: Academic Ds 9B on RTX 5070 Ti report
RTX 50808-bit8K285 tok/s0.1 sPDF: Academic Ds 9B on RTX 5080 report
Radeon RX 7900 XTX8-bit8K226 tok/s0.4 sPDF: Academic Ds 9B on Radeon RX 7900 XTX report
RTX 30908-bit8K279 tok/s0.2 sPDF: Academic Ds 9B on RTX 3090 report
RTX 40908-bit8K295 tok/s0.1 sPDF: Academic Ds 9B on RTX 4090 report
Radeon AI PRO R97008-bit8K169 tok/s0.2 sPDF: Academic Ds 9B on Radeon AI PRO R9700 report
RTX 50908-bit8K429 tok/s0.1 sPDF: Academic Ds 9B on RTX 5090 report
2× RTX 30908-bit8K279 tok/s0.1 sPDF: Academic Ds 9B on 2× RTX 3090 report
2× RTX 40908-bit8K295 tok/s0.1 sPDF: Academic Ds 9B on 2× RTX 4090 report
RTX PRO 5000 Blackwell 48 GB8-bit8K359 tok/s0.1 sPDF: Academic Ds 9B on RTX PRO 5000 Blackwell 48 GB report
RTX 6000 Ada8-bit8K285 tok/s0.1 sPDF: Academic Ds 9B on RTX 6000 Ada report
L40S8-bit8K263 tok/s0.1 sPDF: Academic Ds 9B on L40S report
2× RTX 50908-bit8K429 tok/s0.1 sPDF: Academic Ds 9B on 2× RTX 5090 report
4× RTX 30908-bit8K279 tok/s0.1 sPDF: Academic Ds 9B on 4× RTX 3090 report
RTX PRO 6000 Blackwell8-bit8K429 tok/s0.1 sPDF: Academic Ds 9B on RTX PRO 6000 Blackwell report
Mac mini M4 16 GB8-bit8K51 tok/s5.3 sPDF: Academic Ds 9B on Mac mini M4 16 GB report
Mac mini M4 24 GB8-bit8K51 tok/s5.3 sPDF: Academic Ds 9B on Mac mini M4 24 GB report
Mac mini M4 32 GB8-bit8K51 tok/s5.3 sPDF: Academic Ds 9B on Mac mini M4 32 GB report
Mac mini M4 Pro 24 GB8-bit8K101 tok/s2.6 sPDF: Academic Ds 9B on Mac mini M4 Pro 24 GB report
Mac mini M4 Pro 48 GB8-bit8K101 tok/s2.6 sPDF: Academic Ds 9B on Mac mini M4 Pro 48 GB report
Mac mini M4 Pro 64 GB8-bit8K101 tok/s2.6 sPDF: Academic Ds 9B on Mac mini M4 Pro 64 GB report
M4 Max 128 GB8-bit8K155 tok/s1.3 sPDF: Academic Ds 9B on M4 Max 128 GB report
M5 Max 128 GB8-bit8K168 tok/s0.4 sPDF: Academic Ds 9B on M5 Max 128 GB report
M3 Ultra 96 GB8-bit8K183 tok/s0.7 sPDF: Academic Ds 9B on M3 Ultra 96 GB report
M3 Ultra 256 GB8-bit8K183 tok/s0.7 sPDF: Academic Ds 9B on M3 Ultra 256 GB report
M3 Ultra 512 GB8-bit8K183 tok/s0.7 sPDF: Academic Ds 9B on M3 Ultra 512 GB report
M5 Ultra 512 GB8-bit8K225 tok/s0.2 sPDF: Academic Ds 9B on M5 Ultra 512 GB report
Ryzen AI Max+ 395 64 GB8-bit8K79 tok/s0.8 sPDF: Academic Ds 9B on Ryzen AI Max+ 395 64 GB report
Ryzen AI Max+ 395 128 GB8-bit8K79 tok/s0.8 sPDF: Academic Ds 9B on Ryzen AI Max+ 395 128 GB report
RTX Spark8-bit8K109 tok/s0.3 sPDF: Academic Ds 9B on RTX Spark report
DGX Spark8-bit8K101 tok/s0.3 sPDF: Academic Ds 9B on DGX Spark report
Dual DGX Spark8-bit8K101 tok/s0.2 sPDF: Academic Ds 9B on Dual DGX Spark report
DGX Station GB3008-bit8K766 tok/s0.1 sPDF: Academic Ds 9B on DGX Station GB300 report
2× RTX PRO 6000 Blackwell8-bit8K429 tok/s0.1 sPDF: Academic Ds 9B on 2× RTX PRO 6000 Blackwell report
4× RTX PRO 6000 Blackwell8-bit8K429 tok/s0.1 sPDF: Academic Ds 9B on 4× RTX PRO 6000 Blackwell report
8× RTX PRO 6000 Blackwell8-bit8K429 tok/s0.1 sPDF: Academic Ds 9B on 8× RTX PRO 6000 Blackwell report
A100 80 GB SXM8-bit8K462 tok/s0.2 sPDF: Academic Ds 9B on A100 80 GB SXM report
H100 SXM8-bit8K591 tok/s0.1 sPDF: Academic Ds 9B on H100 SXM report
4× H200 NVL8-bit8K680 tok/s0.1 sPDF: Academic Ds 9B on 4× H200 NVL report
8× H200 NVL8-bit8K680 tok/s0.1 sPDF: Academic Ds 9B 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/academic-ds-9B-GGUF
vLLM
vllm serve ByteDance-Seed/academic-ds-9B

Questions

How much VRAM does Academic Ds 9B need?

Academic Ds 9B needs 5.3 GB for its weights at 4-bit, 6.2 GB with an 8K context and 6.2 GB with 8K. At 8-bit, 10.2 GB.

What is the minimum VRAM for Academic Ds 9B?

About 5.2 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 Academic Ds 9B?

9.4B parameters (1.6B active per token): 17.4 GB of weights at 16-bit, 5.3 GB at 4-bit.

Can Academic Ds 9B run on an RTX 4090?

Yes. At 8-bit it holds 8K of context and writes about 295 tokens a second.

More models