Skip to content

Gemma SEA LION V3 9B IT VRAM requirements

How much VRAM Gemma SEA LION V3 9B IT needs
4-bit weights: 5.2 GBWith 8K context: 8.9 GB8-bit, 8K context: 12.8 GB
Parameters
9.2B
Longest context
8K
Released
Oct 2024
License
MIT
Strongest at
Not rated

How much VRAM Gemma SEA LION V3 9B IT needs

Memory Gemma SEA LION V3 9B IT needs at each precision and context length
PrecisionWeightsWith 8K
16-bit17.2 GB20.9 GB
8-bit9.1 GB12.8 GB
6-bit7.1 GB10.7 GB
5-bit6.1 GB9.8 GB
4-bit5.2 GB8.9 GB
3-bit4.2 GB7.9 GB
2-bit3.2 GB6.9 GB
1.6-bit1.8 GB5.4 GB

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

GPUs and Macs that run Gemma SEA LION V3 9B IT

The best way Gemma SEA LION V3 9B IT runs on common GPUs, Macs and AI boxes
MachineBest fitContextSpeedFirst replyReport
RTX 4060Too bigPDF: Gemma SEA LION V3 9B IT on RTX 4060 report
RTX 5060Too bigPDF: Gemma SEA LION V3 9B IT on RTX 5060 report
RTX 3070Too bigPDF: Gemma SEA LION V3 9B IT on RTX 3070 report
RTX 3080 10 GB4-bit8K93 tok/s0.6 sPDF: Gemma SEA LION V3 9B IT on RTX 3080 10 GB report
RTX 3060 12 GB6-bit8K35 tok/s1.3 sPDF: Gemma SEA LION V3 9B IT on RTX 3060 12 GB report
Arc B5806-bit8K31 tok/s0.9 sPDF: Gemma SEA LION V3 9B IT on Arc B580 report
RTX 40706-bit8K49 tok/s0.6 sPDF: Gemma SEA LION V3 9B IT on RTX 4070 report
RTX 4070 Super6-bit8K49 tok/s0.5 sPDF: Gemma SEA LION V3 9B IT on RTX 4070 Super report
RTX 50706-bit8K63 tok/s0.5 sPDF: Gemma SEA LION V3 9B IT on RTX 5070 report
RTX 4060 Ti 16 GB8-bit8K22 tok/s0.8 sPDF: Gemma SEA LION V3 9B IT on RTX 4060 Ti 16 GB report
RTX 5060 Ti 16 GB8-bit8K34 tok/s0.7 sPDF: Gemma SEA LION V3 9B IT on RTX 5060 Ti 16 GB report
Arc A770 16 GB8-bit8K29 tok/s0.8 sPDF: Gemma SEA LION V3 9B IT on Arc A770 16 GB report
Radeon RX 9070 XT8-bit8K40 tok/s0.5 sPDF: Gemma SEA LION V3 9B IT on Radeon RX 9070 XT report
RTX 4070 Ti Super8-bit8K50 tok/s0.4 sPDF: Gemma SEA LION V3 9B IT on RTX 4070 Ti Super report
RTX 4080 Super8-bit8K54 tok/s0.3 sPDF: Gemma SEA LION V3 9B IT on RTX 4080 Super report
RTX 5070 Ti8-bit8K65 tok/s0.4 sPDF: Gemma SEA LION V3 9B IT on RTX 5070 Ti report
RTX 50808-bit8K69 tok/s0.3 sPDF: Gemma SEA LION V3 9B IT on RTX 5080 report
Radeon RX 7900 XTX8-bit8K57 tok/s0.7 sPDF: Gemma SEA LION V3 9B IT on Radeon RX 7900 XTX report
RTX 30908-bit8K68 tok/s0.5 sPDF: Gemma SEA LION V3 9B IT on RTX 3090 report
RTX 40908-bit8K72 tok/s0.2 sPDF: Gemma SEA LION V3 9B IT on RTX 4090 report
Radeon AI PRO R97008-bit8K40 tok/s0.5 sPDF: Gemma SEA LION V3 9B IT on Radeon AI PRO R9700 report
RTX 50908-bit8K120 tok/s0.2 sPDF: Gemma SEA LION V3 9B IT on RTX 5090 report
2× RTX 30908-bit8K68 tok/s0.3 sPDF: Gemma SEA LION V3 9B IT on 2× RTX 3090 report
2× RTX 40908-bit8K72 tok/s0.1 sPDF: Gemma SEA LION V3 9B IT on 2× RTX 4090 report
RTX PRO 5000 Blackwell 48 GB8-bit8K94 tok/s0.2 sPDF: Gemma SEA LION V3 9B IT on RTX PRO 5000 Blackwell 48 GB report
RTX 6000 Ada8-bit8K69 tok/s0.2 sPDF: Gemma SEA LION V3 9B IT on RTX 6000 Ada report
L40S8-bit8K63 tok/s0.3 sPDF: Gemma SEA LION V3 9B IT on L40S report
2× RTX 50908-bit8K120 tok/s0.1 sPDF: Gemma SEA LION V3 9B IT on 2× RTX 5090 report
4× RTX 30908-bit8K68 tok/s0.1 sPDF: Gemma SEA LION V3 9B IT on 4× RTX 3090 report
RTX PRO 6000 Blackwell8-bit8K120 tok/s0.1 sPDF: Gemma SEA LION V3 9B IT on RTX PRO 6000 Blackwell report
Mac mini M4 16 GB5-bit8K16 tok/s11 sPDF: Gemma SEA LION V3 9B IT on Mac mini M4 16 GB report
Mac mini M4 24 GB8-bit8K11 tok/s11 sPDF: Gemma SEA LION V3 9B IT on Mac mini M4 24 GB report
Mac mini M4 32 GB8-bit8K11 tok/s11 sPDF: Gemma SEA LION V3 9B IT on Mac mini M4 32 GB report
Mac mini M4 Pro 24 GB8-bit8K23 tok/s5.3 sPDF: Gemma SEA LION V3 9B IT on Mac mini M4 Pro 24 GB report
Mac mini M4 Pro 48 GB8-bit8K23 tok/s5.3 sPDF: Gemma SEA LION V3 9B IT on Mac mini M4 Pro 48 GB report
Mac mini M4 Pro 64 GB8-bit8K23 tok/s5.3 sPDF: Gemma SEA LION V3 9B IT on Mac mini M4 Pro 64 GB report
M4 Max 128 GB8-bit8K40 tok/s2.6 sPDF: Gemma SEA LION V3 9B IT on M4 Max 128 GB report
M5 Max 128 GB8-bit8K44 tok/s0.8 sPDF: Gemma SEA LION V3 9B IT on M5 Max 128 GB report
M3 Ultra 96 GB8-bit8K50 tok/s1.5 sPDF: Gemma SEA LION V3 9B IT on M3 Ultra 96 GB report
M3 Ultra 256 GB8-bit8K50 tok/s1.5 sPDF: Gemma SEA LION V3 9B IT on M3 Ultra 256 GB report
M3 Ultra 512 GB8-bit8K50 tok/s1.5 sPDF: Gemma SEA LION V3 9B IT on M3 Ultra 512 GB report
M5 Ultra 512 GB8-bit8K68 tok/s0.5 sPDF: Gemma SEA LION V3 9B IT on M5 Ultra 512 GB report
Ryzen AI Max+ 395 64 GB8-bit8K17 tok/s1.5 sPDF: Gemma SEA LION V3 9B IT on Ryzen AI Max+ 395 64 GB report
Ryzen AI Max+ 395 128 GB8-bit8K17 tok/s1.5 sPDF: Gemma SEA LION V3 9B IT on Ryzen AI Max+ 395 128 GB report
RTX Spark8-bit8K23 tok/s0.5 sPDF: Gemma SEA LION V3 9B IT on RTX Spark report
DGX Spark8-bit8K21 tok/s0.5 sPDF: Gemma SEA LION V3 9B IT on DGX Spark report
Dual DGX Spark8-bit8K21 tok/s0.3 sPDF: Gemma SEA LION V3 9B IT on Dual DGX Spark report
DGX Station GB3008-bit8K329 tok/s0.1 sPDF: Gemma SEA LION V3 9B IT on DGX Station GB300 report
2× RTX PRO 6000 Blackwell8-bit8K120 tok/s0.1 sPDF: Gemma SEA LION V3 9B IT on 2× RTX PRO 6000 Blackwell report
4× RTX PRO 6000 Blackwell8-bit8K120 tok/s0.1 sPDF: Gemma SEA LION V3 9B IT on 4× RTX PRO 6000 Blackwell report
8× RTX PRO 6000 Blackwell8-bit8K120 tok/s0.1 sPDF: Gemma SEA LION V3 9B IT on 8× RTX PRO 6000 Blackwell report
A100 80 GB SXM8-bit8K134 tok/s0.3 sPDF: Gemma SEA LION V3 9B IT on A100 80 GB SXM report
H100 SXM8-bit8K199 tok/s0.1 sPDF: Gemma SEA LION V3 9B IT on H100 SXM report
4× H200 NVL8-bit8K257 tok/s0.1 sPDF: Gemma SEA LION V3 9B IT on 4× H200 NVL report
8× H200 NVL8-bit8K257 tok/s0.1 sPDF: Gemma SEA LION V3 9B IT 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 aisingapore/Gemma-SEA-LION-v3-9B-IT-GGUF
vLLM
vllm serve aisingapore/Gemma-SEA-LION-v3-9B-IT

Questions

How much VRAM does Gemma SEA LION V3 9B IT need?

Gemma SEA LION V3 9B IT needs 5.2 GB for its weights at 4-bit, 8.9 GB with an 8K context and 8.9 GB with 8K. At 8-bit, 12.8 GB.

What is the minimum VRAM for Gemma SEA LION V3 9B IT?

About 7.9 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 3080 10 GB.

How big is Gemma SEA LION V3 9B IT?

9.2B parameters: 17.2 GB of weights at 16-bit, 5.2 GB at 4-bit.

Can Gemma SEA LION V3 9B IT run on an RTX 4090?

Yes. At 8-bit it holds 8K of context and writes about 72 tokens a second.

More models