Skip to content

Inkling Small VRAM requirements

How much VRAM Inkling Small needs
4-bit weights: 150 GBWith 4K context: 152 GB8-bit, 4K context: 265 GB
Parameters
266B
Active per token
11.4B
Longest context
4K
Released
Jul 2026
License
Apache 2.0
Strongest at
Text and images

How much VRAM Inkling Small needs

Memory Inkling Small needs at each precision and context length
PrecisionWeightsWith 4K
16-bit495 GB497 GB
8-bit263 GB265 GB
6-bit203 GB205 GB
5-bit176 GB178 GB
4-bit150 GB152 GB
3-bit121 GB123 GB
2-bit92.9 GB94.5 GB
1.6-bit51.1 GB52.7 GB

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

GPUs and Macs that run Inkling Small

The best way Inkling Small runs on common GPUs, Macs and AI boxes
MachineBest fitContextSpeedFirst replyReport
RTX 4060Too bigPDF: Inkling Small on RTX 4060 report
RTX 5060Too bigPDF: Inkling Small on RTX 5060 report
RTX 3070Too bigPDF: Inkling Small on RTX 3070 report
RTX 3080 10 GBToo bigPDF: Inkling Small on RTX 3080 10 GB report
RTX 3060 12 GBToo bigPDF: Inkling Small on RTX 3060 12 GB report
Arc B580Too bigPDF: Inkling Small on Arc B580 report
RTX 4070Too bigPDF: Inkling Small on RTX 4070 report
RTX 4070 SuperToo bigPDF: Inkling Small on RTX 4070 Super report
RTX 5070Too bigPDF: Inkling Small on RTX 5070 report
RTX 4060 Ti 16 GBToo bigPDF: Inkling Small on RTX 4060 Ti 16 GB report
RTX 5060 Ti 16 GBToo bigPDF: Inkling Small on RTX 5060 Ti 16 GB report
Arc A770 16 GBToo bigPDF: Inkling Small on Arc A770 16 GB report
Radeon RX 9070 XTToo bigPDF: Inkling Small on Radeon RX 9070 XT report
RTX 4070 Ti SuperToo bigPDF: Inkling Small on RTX 4070 Ti Super report
RTX 4080 SuperToo bigPDF: Inkling Small on RTX 4080 Super report
RTX 5070 TiToo bigPDF: Inkling Small on RTX 5070 Ti report
RTX 5080Too bigPDF: Inkling Small on RTX 5080 report
Radeon RX 7900 XTXToo bigPDF: Inkling Small on Radeon RX 7900 XTX report
RTX 3090Too bigPDF: Inkling Small on RTX 3090 report
RTX 4090Too bigPDF: Inkling Small on RTX 4090 report
Radeon AI PRO R9700Too bigPDF: Inkling Small on Radeon AI PRO R9700 report
RTX 5090Too bigPDF: Inkling Small on RTX 5090 report
2× RTX 3090Too bigPDF: Inkling Small on 2× RTX 3090 report
2× RTX 4090Too bigPDF: Inkling Small on 2× RTX 4090 report
RTX PRO 5000 Blackwell 48 GBToo bigPDF: Inkling Small on RTX PRO 5000 Blackwell 48 GB report
RTX 6000 AdaToo bigPDF: Inkling Small on RTX 6000 Ada report
L40SToo bigPDF: Inkling Small on L40S report
2× RTX 5090Too bigPDF: Inkling Small on 2× RTX 5090 report
4× RTX 3090Too bigPDF: Inkling Small on 4× RTX 3090 report
RTX PRO 6000 Blackwell2-bit4K168 tok/s0.6 sPDF: Inkling Small on RTX PRO 6000 Blackwell report
Mac mini M4 16 GBToo bigPDF: Inkling Small on Mac mini M4 16 GB report
Mac mini M4 24 GBToo bigPDF: Inkling Small on Mac mini M4 24 GB report
Mac mini M4 32 GBToo bigPDF: Inkling Small on Mac mini M4 32 GB report
Mac mini M4 Pro 24 GBToo bigPDF: Inkling Small on Mac mini M4 Pro 24 GB report
Mac mini M4 Pro 48 GBToo bigPDF: Inkling Small on Mac mini M4 Pro 48 GB report
Mac mini M4 Pro 64 GBToo bigPDF: Inkling Small on Mac mini M4 Pro 64 GB report
M4 Max 128 GB2-bit4K61 tok/s12 sPDF: Inkling Small on M4 Max 128 GB report
M5 Max 128 GB2-bit4K66 tok/s3.3 sPDF: Inkling Small on M5 Max 128 GB report
M3 Ultra 96 GBToo bigPDF: Inkling Small on M3 Ultra 96 GB report
M3 Ultra 256 GB5-bit4K47 tok/s6.5 sPDF: Inkling Small on M3 Ultra 256 GB report
M3 Ultra 512 GB8-bit4K35 tok/s6.5 sPDF: Inkling Small on M3 Ultra 512 GB report
M5 Ultra 512 GB8-bit4K47 tok/s2.0 sPDF: Inkling Small on M5 Ultra 512 GB report
Ryzen AI Max+ 395 64 GBToo bigPDF: Inkling Small on Ryzen AI Max+ 395 64 GB report
Ryzen AI Max+ 395 128 GB2-bit4K32 tok/s6.7 sPDF: Inkling Small on Ryzen AI Max+ 395 128 GB report
RTX Spark2-bit4K44 tok/s2.3 sPDF: Inkling Small on RTX Spark report
DGX Spark2-bit4K40 tok/s2.3 sPDF: Inkling Small on DGX Spark report
Dual DGX Spark6-bit4K19 tok/s2.6 sPDF: Inkling Small on Dual DGX Spark report
DGX Station GB3006-bit4K227 tok/s0.2 sPDF: Inkling Small on DGX Station GB300 report
2× RTX PRO 6000 Blackwell5-bit4K111 tok/s0.3 sPDF: Inkling Small on 2× RTX PRO 6000 Blackwell report
4× RTX PRO 6000 Blackwell8-bit4K82 tok/s0.3 sPDF: Inkling Small on 4× RTX PRO 6000 Blackwell report
8× RTX PRO 6000 Blackwell8-bit4K82 tok/s0.1 sPDF: Inkling Small on 8× RTX PRO 6000 Blackwell report
A100 80 GB SXMToo bigPDF: Inkling Small on A100 80 GB SXM report
H100 SXMToo bigPDF: Inkling Small on H100 SXM report
4× H200 NVL8-bit4K163 tok/s0.2 sPDF: Inkling Small on 4× H200 NVL report
8× H200 NVL8-bit4K163 tok/s0.1 sPDF: Inkling Small 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/Inkling-Small-GGUF
vLLM
vllm serve thinkingmachines/Inkling-Small
MLX (Mac)
mlx_lm.chat --model mlx-community/Inkling-Small-mxfp4

Questions

How much VRAM does Inkling Small need?

Inkling Small needs 150 GB for its weights at 4-bit, 152 GB with an 4K context and 152 GB with 4K. At 8-bit, 265 GB.

What is the minimum VRAM for Inkling Small?

About 94.5 GB at 2-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 PRO 6000 Blackwell.

How big is Inkling Small?

266B parameters (11.4B active per token): 495 GB of weights at 16-bit, 150 GB at 4-bit.

Can Inkling Small run on an RTX 4090?

Not on one card. The smallest machine here that runs it is the RTX PRO 6000 Blackwell.

More models