Skip to content

Pythia 1.4B Deduped VRAM requirements

How much VRAM Pythia 1.4B Deduped needs
4-bit weights: 818 MBWith 2K context: 1.8 GB8-bit, 2K context: 2.4 GB
Parameters
1.4B
Longest context
2K
Released
Feb 2023
License
Apache 2.0
Strongest at
Not rated

How much VRAM Pythia 1.4B Deduped needs

Memory Pythia 1.4B Deduped needs at each precision and context length
PrecisionWeightsWith 2K
16-bit2.6 GB3.6 GB
8-bit1.4 GB2.4 GB
6-bit1.1 GB2.1 GB
5-bit960 MB1.9 GB
4-bit818 MB1.8 GB
3-bit659 MB1.6 GB
2-bit506 MB1.5 GB
1.6-bit278 MB1.3 GB

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

GPUs and Macs that run Pythia 1.4B Deduped

The best way Pythia 1.4B Deduped runs on common GPUs, Macs and AI boxes
MachineBest fitContextSpeedFirst replyReport
RTX 40608-bit2K124 tok/s0.2 sPDF: Pythia 1.4B Deduped on RTX 4060 report
RTX 50608-bit2K194 tok/s0.1 sPDF: Pythia 1.4B Deduped on RTX 5060 report
RTX 30708-bit2K194 tok/s0.1 sPDF: Pythia 1.4B Deduped on RTX 3070 report
RTX 3080 10 GB8-bit2K299 tok/s0.1 sPDF: Pythia 1.4B Deduped on RTX 3080 10 GB report
RTX 3060 12 GB8-bit2K160 tok/s0.2 sPDF: Pythia 1.4B Deduped on RTX 3060 12 GB report
Arc B5808-bit2K132 tok/s0.2 sPDF: Pythia 1.4B Deduped on Arc B580 report
RTX 40708-bit2K214 tok/s0.1 sPDF: Pythia 1.4B Deduped on RTX 4070 report
RTX 4070 Super8-bit2K214 tok/s0.1 sPDF: Pythia 1.4B Deduped on RTX 4070 Super report
RTX 50708-bit2K272 tok/s0.1 sPDF: Pythia 1.4B Deduped on RTX 5070 report
RTX 4060 Ti 16 GB8-bit2K131 tok/s0.1 sPDF: Pythia 1.4B Deduped on RTX 4060 Ti 16 GB report
RTX 5060 Ti 16 GB8-bit2K194 tok/s0.1 sPDF: Pythia 1.4B Deduped on RTX 5060 Ti 16 GB report
Arc A770 16 GB8-bit2K156 tok/s0.1 sPDF: Pythia 1.4B Deduped on Arc A770 16 GB report
Radeon RX 9070 XT8-bit2K207 tok/s0.1 sPDF: Pythia 1.4B Deduped on Radeon RX 9070 XT report
RTX 4070 Ti Super8-bit2K272 tok/s0.1 sPDF: Pythia 1.4B Deduped on RTX 4070 Ti Super report
RTX 4080 Super8-bit2K292 tok/s0.1 sPDF: Pythia 1.4B Deduped on RTX 4080 Super report
RTX 5070 Ti8-bit2K340 tok/s0.1 sPDF: Pythia 1.4B Deduped on RTX 5070 Ti report
RTX 50808-bit2K358 tok/s0.1 sPDF: Pythia 1.4B Deduped on RTX 5080 report
Radeon RX 7900 XTX8-bit2K276 tok/s0.1 sPDF: Pythia 1.4B Deduped on Radeon RX 7900 XTX report
RTX 30908-bit2K351 tok/s0.1 sPDF: Pythia 1.4B Deduped on RTX 3090 report
RTX 40908-bit2K371 tok/s0.1 sPDF: Pythia 1.4B Deduped on RTX 4090 report
Radeon AI PRO R97008-bit2K207 tok/s0.1 sPDF: Pythia 1.4B Deduped on Radeon AI PRO R9700 report
RTX 50908-bit2K546 tok/s0.1 sPDF: Pythia 1.4B Deduped on RTX 5090 report
2× RTX 30908-bit2K351 tok/s0.1 sPDF: Pythia 1.4B Deduped on 2× RTX 3090 report
2× RTX 40908-bit2K371 tok/s0.1 sPDF: Pythia 1.4B Deduped on 2× RTX 4090 report
RTX PRO 5000 Blackwell 48 GB8-bit2K454 tok/s0.1 sPDF: Pythia 1.4B Deduped on RTX PRO 5000 Blackwell 48 GB report
RTX 6000 Ada8-bit2K358 tok/s0.1 sPDF: Pythia 1.4B Deduped on RTX 6000 Ada report
L40S8-bit2K331 tok/s0.1 sPDF: Pythia 1.4B Deduped on L40S report
2× RTX 50908-bit2K546 tok/s0.1 sPDF: Pythia 1.4B Deduped on 2× RTX 5090 report
4× RTX 30908-bit2K351 tok/s0.1 sPDF: Pythia 1.4B Deduped on 4× RTX 3090 report
RTX PRO 6000 Blackwell8-bit2K546 tok/s0.1 sPDF: Pythia 1.4B Deduped on RTX PRO 6000 Blackwell report
Mac mini M4 16 GB8-bit2K62 tok/s1.8 sPDF: Pythia 1.4B Deduped on Mac mini M4 16 GB report
Mac mini M4 24 GB8-bit2K62 tok/s1.8 sPDF: Pythia 1.4B Deduped on Mac mini M4 24 GB report
Mac mini M4 32 GB8-bit2K62 tok/s1.8 sPDF: Pythia 1.4B Deduped on Mac mini M4 32 GB report
Mac mini M4 Pro 24 GB8-bit2K123 tok/s0.9 sPDF: Pythia 1.4B Deduped on Mac mini M4 Pro 24 GB report
Mac mini M4 Pro 48 GB8-bit2K123 tok/s0.9 sPDF: Pythia 1.4B Deduped on Mac mini M4 Pro 48 GB report
Mac mini M4 Pro 64 GB8-bit2K123 tok/s0.9 sPDF: Pythia 1.4B Deduped on Mac mini M4 Pro 64 GB report
M4 Max 128 GB8-bit2K188 tok/s0.4 sPDF: Pythia 1.4B Deduped on M4 Max 128 GB report
M5 Max 128 GB8-bit2K202 tok/s0.1 sPDF: Pythia 1.4B Deduped on M5 Max 128 GB report
M3 Ultra 96 GB8-bit2K220 tok/s0.3 sPDF: Pythia 1.4B Deduped on M3 Ultra 96 GB report
M3 Ultra 256 GB8-bit2K220 tok/s0.3 sPDF: Pythia 1.4B Deduped on M3 Ultra 256 GB report
M3 Ultra 512 GB8-bit2K220 tok/s0.3 sPDF: Pythia 1.4B Deduped on M3 Ultra 512 GB report
M5 Ultra 512 GB8-bit2K270 tok/s0.1 sPDF: Pythia 1.4B Deduped on M5 Ultra 512 GB report
Ryzen AI Max+ 395 64 GB8-bit2K97 tok/s0.3 sPDF: Pythia 1.4B Deduped on Ryzen AI Max+ 395 64 GB report
Ryzen AI Max+ 395 128 GB8-bit2K97 tok/s0.3 sPDF: Pythia 1.4B Deduped on Ryzen AI Max+ 395 128 GB report
RTX Spark8-bit2K136 tok/s0.1 sPDF: Pythia 1.4B Deduped on RTX Spark report
DGX Spark8-bit2K125 tok/s0.1 sPDF: Pythia 1.4B Deduped on DGX Spark report
Dual DGX Spark8-bit2K125 tok/s0.1 sPDF: Pythia 1.4B Deduped on Dual DGX Spark report
DGX Station GB3008-bit2K1,000 tok/s0.1 sPDF: Pythia 1.4B Deduped on DGX Station GB300 report
2× RTX PRO 6000 Blackwell8-bit2K546 tok/s0.1 sPDF: Pythia 1.4B Deduped on 2× RTX PRO 6000 Blackwell report
4× RTX PRO 6000 Blackwell8-bit2K546 tok/s0.1 sPDF: Pythia 1.4B Deduped on 4× RTX PRO 6000 Blackwell report
8× RTX PRO 6000 Blackwell8-bit2K546 tok/s0.1 sPDF: Pythia 1.4B Deduped on 8× RTX PRO 6000 Blackwell report
A100 80 GB SXM8-bit2K589 tok/s0.1 sPDF: Pythia 1.4B Deduped on A100 80 GB SXM report
H100 SXM8-bit2K761 tok/s0.1 sPDF: Pythia 1.4B Deduped on H100 SXM report
4× H200 NVL8-bit2K881 tok/s0.1 sPDF: Pythia 1.4B Deduped on 4× H200 NVL report
8× H200 NVL8-bit2K881 tok/s0.1 sPDF: Pythia 1.4B Deduped 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/pythia-1.4b-deduped-i1-GGUF
vLLM
vllm serve EleutherAI/pythia-1.4b-deduped

Questions

How much VRAM does Pythia 1.4B Deduped need?

Pythia 1.4B Deduped needs 818 MB for its weights at 4-bit, 1.8 GB with an 2K context and 1.8 GB with 2K. At 8-bit, 2.4 GB.

What is the minimum VRAM for Pythia 1.4B Deduped?

About 1.6 GB at 3-bit with an 2K 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 Pythia 1.4B Deduped?

1.4B parameters: 2.6 GB of weights at 16-bit, 818 MB at 4-bit.

Can Pythia 1.4B Deduped run on an RTX 4090?

Yes. At 8-bit it holds 2K of context and writes about 371 tokens a second.

More models