Skip to content

Yi 1.5 6B VRAM requirements

How much VRAM Yi 1.5 6B needs
4-bit weights: 3.4 GBWith 4K context: 4.4 GB8-bit, 4K context: 6.9 GB
Parameters
6.1B
Longest context
4K
Released
May 2024
License
Apache 2.0
Strongest at
Not rated

How much VRAM Yi 1.5 6B needs

Memory Yi 1.5 6B needs at each precision and context length
PrecisionWeightsWith 4K
16-bit11.3 GB12.2 GB
8-bit6.0 GB6.9 GB
6-bit4.6 GB5.6 GB
5-bit4.0 GB5.0 GB
4-bit3.4 GB4.4 GB
3-bit2.8 GB3.7 GB
2-bit2.1 GB3.1 GB
1.6-bit1.2 GB2.1 GB

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

GPUs and Macs that run Yi 1.5 6B

The best way Yi 1.5 6B runs on common GPUs, Macs and AI boxes
MachineBest fitContextSpeedFirst replyReport
RTX 40608-bit4K30 tok/s0.8 sPDF: Yi 1.5 6B on RTX 4060 report
RTX 50608-bit4K49 tok/s0.7 sPDF: Yi 1.5 6B on RTX 5060 report
RTX 30708-bit4K49 tok/s0.6 sPDF: Yi 1.5 6B on RTX 3070 report
RTX 3080 10 GB8-bit4K80 tok/s0.4 sPDF: Yi 1.5 6B on RTX 3080 10 GB report
RTX 3060 12 GB8-bit4K39 tok/s1.0 sPDF: Yi 1.5 6B on RTX 3060 12 GB report
Arc B5808-bit4K35 tok/s0.7 sPDF: Yi 1.5 6B on Arc B580 report
RTX 40708-bit4K54 tok/s0.4 sPDF: Yi 1.5 6B on RTX 4070 report
RTX 4070 Super8-bit4K54 tok/s0.4 sPDF: Yi 1.5 6B on RTX 4070 Super report
RTX 50708-bit4K71 tok/s0.4 sPDF: Yi 1.5 6B on RTX 5070 report
RTX 4060 Ti 16 GB8-bit4K32 tok/s0.6 sPDF: Yi 1.5 6B on RTX 4060 Ti 16 GB report
RTX 5060 Ti 16 GB8-bit4K49 tok/s0.5 sPDF: Yi 1.5 6B on RTX 5060 Ti 16 GB report
Arc A770 16 GB8-bit4K42 tok/s0.6 sPDF: Yi 1.5 6B on Arc A770 16 GB report
Radeon RX 9070 XT8-bit4K57 tok/s0.3 sPDF: Yi 1.5 6B on Radeon RX 9070 XT report
RTX 4070 Ti Super8-bit4K71 tok/s0.3 sPDF: Yi 1.5 6B on RTX 4070 Ti Super report
RTX 4080 Super8-bit4K77 tok/s0.2 sPDF: Yi 1.5 6B on RTX 4080 Super report
RTX 5070 Ti8-bit4K93 tok/s0.3 sPDF: Yi 1.5 6B on RTX 5070 Ti report
RTX 50808-bit4K99 tok/s0.2 sPDF: Yi 1.5 6B on RTX 5080 report
Radeon RX 7900 XTX8-bit4K81 tok/s0.6 sPDF: Yi 1.5 6B on Radeon RX 7900 XTX report
RTX 30908-bit4K96 tok/s0.4 sPDF: Yi 1.5 6B on RTX 3090 report
RTX 40908-bit4K103 tok/s0.2 sPDF: Yi 1.5 6B on RTX 4090 report
Radeon AI PRO R97008-bit4K57 tok/s0.4 sPDF: Yi 1.5 6B on Radeon AI PRO R9700 report
RTX 50908-bit4K170 tok/s0.1 sPDF: Yi 1.5 6B on RTX 5090 report
2× RTX 30908-bit4K96 tok/s0.2 sPDF: Yi 1.5 6B on 2× RTX 3090 report
2× RTX 40908-bit4K103 tok/s0.1 sPDF: Yi 1.5 6B on 2× RTX 4090 report
RTX PRO 5000 Blackwell 48 GB8-bit4K133 tok/s0.2 sPDF: Yi 1.5 6B on RTX PRO 5000 Blackwell 48 GB report
RTX 6000 Ada8-bit4K99 tok/s0.1 sPDF: Yi 1.5 6B on RTX 6000 Ada report
L40S8-bit4K90 tok/s0.2 sPDF: Yi 1.5 6B on L40S report
2× RTX 50908-bit4K170 tok/s0.1 sPDF: Yi 1.5 6B on 2× RTX 5090 report
4× RTX 30908-bit4K96 tok/s0.1 sPDF: Yi 1.5 6B on 4× RTX 3090 report
RTX PRO 6000 Blackwell8-bit4K170 tok/s0.1 sPDF: Yi 1.5 6B on RTX PRO 6000 Blackwell report
Mac mini M4 16 GB8-bit4K15 tok/s8.0 sPDF: Yi 1.5 6B on Mac mini M4 16 GB report
Mac mini M4 24 GB8-bit4K15 tok/s8.0 sPDF: Yi 1.5 6B on Mac mini M4 24 GB report
Mac mini M4 32 GB8-bit4K15 tok/s8.0 sPDF: Yi 1.5 6B on Mac mini M4 32 GB report
Mac mini M4 Pro 24 GB8-bit4K33 tok/s4.0 sPDF: Yi 1.5 6B on Mac mini M4 Pro 24 GB report
Mac mini M4 Pro 48 GB8-bit4K33 tok/s4.0 sPDF: Yi 1.5 6B on Mac mini M4 Pro 48 GB report
Mac mini M4 Pro 64 GB8-bit4K33 tok/s4.0 sPDF: Yi 1.5 6B on Mac mini M4 Pro 64 GB report
M4 Max 128 GB8-bit4K57 tok/s2.0 sPDF: Yi 1.5 6B on M4 Max 128 GB report
M5 Max 128 GB8-bit4K63 tok/s0.6 sPDF: Yi 1.5 6B on M5 Max 128 GB report
M3 Ultra 96 GB8-bit4K71 tok/s1.1 sPDF: Yi 1.5 6B on M3 Ultra 96 GB report
M3 Ultra 256 GB8-bit4K71 tok/s1.1 sPDF: Yi 1.5 6B on M3 Ultra 256 GB report
M3 Ultra 512 GB8-bit4K71 tok/s1.1 sPDF: Yi 1.5 6B on M3 Ultra 512 GB report
M5 Ultra 512 GB8-bit4K96 tok/s0.3 sPDF: Yi 1.5 6B on M5 Ultra 512 GB report
Ryzen AI Max+ 395 64 GB8-bit4K24 tok/s1.1 sPDF: Yi 1.5 6B on Ryzen AI Max+ 395 64 GB report
Ryzen AI Max+ 395 128 GB8-bit4K24 tok/s1.1 sPDF: Yi 1.5 6B on Ryzen AI Max+ 395 128 GB report
RTX Spark8-bit4K33 tok/s0.4 sPDF: Yi 1.5 6B on RTX Spark report
DGX Spark8-bit4K30 tok/s0.4 sPDF: Yi 1.5 6B on DGX Spark report
Dual DGX Spark8-bit4K30 tok/s0.2 sPDF: Yi 1.5 6B on Dual DGX Spark report
DGX Station GB3008-bit4K455 tok/s0.1 sPDF: Yi 1.5 6B on DGX Station GB300 report
2× RTX PRO 6000 Blackwell8-bit4K170 tok/s0.1 sPDF: Yi 1.5 6B on 2× RTX PRO 6000 Blackwell report
4× RTX PRO 6000 Blackwell8-bit4K170 tok/s0.1 sPDF: Yi 1.5 6B on 4× RTX PRO 6000 Blackwell report
8× RTX PRO 6000 Blackwell8-bit4K170 tok/s0.1 sPDF: Yi 1.5 6B on 8× RTX PRO 6000 Blackwell report
A100 80 GB SXM8-bit4K190 tok/s0.3 sPDF: Yi 1.5 6B on A100 80 GB SXM report
H100 SXM8-bit4K279 tok/s0.1 sPDF: Yi 1.5 6B on H100 SXM report
4× H200 NVL8-bit4K358 tok/s0.1 sPDF: Yi 1.5 6B on 4× H200 NVL report
8× H200 NVL8-bit4K358 tok/s0.1 sPDF: Yi 1.5 6B 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/Yi-1.5-6B-i1-GGUF
vLLM
vllm serve 01-ai/Yi-1.5-6B

Questions

How much VRAM does Yi 1.5 6B need?

Yi 1.5 6B needs 3.4 GB for its weights at 4-bit, 4.4 GB with an 4K context and 4.4 GB with 4K. At 8-bit, 6.9 GB.

What is the minimum VRAM for Yi 1.5 6B?

About 3.7 GB at 3-bit with an 4K 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 Yi 1.5 6B?

6.1B parameters: 11.3 GB of weights at 16-bit, 3.4 GB at 4-bit.

Can Yi 1.5 6B run on an RTX 4090?

Yes. At 8-bit it holds 4K of context and writes about 103 tokens a second.

More models