Skip to content

Midm 2.0 Base Instruct VRAM requirements

How much VRAM Midm 2.0 Base Instruct needs
4-bit weights: 6.5 GBWith 8K context: 8.8 GBWith 32K context: 13.3 GB8-bit, 8K context: 13.8 GB
Parameters
11.5B
Longest context
32K
Released
Jul 2025
License
MIT
Strongest at
Not rated

How much VRAM Midm 2.0 Base Instruct needs

Memory Midm 2.0 Base Instruct needs at each precision and context length
PrecisionWeightsWith 8KWith 32K
16-bit21.5 GB23.8 GB28.3 GB
8-bit11.4 GB13.8 GB18.3 GB
6-bit8.8 GB11.1 GB15.6 GB
5-bit7.6 GB10.0 GB14.5 GB
4-bit6.5 GB8.8 GB13.3 GB
3-bit5.3 GB7.6 GB12.1 GB
2-bit4.0 GB6.4 GB10.9 GB
1.6-bit2.2 GB4.5 GB9.0 GB

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

GPUs and Macs that run Midm 2.0 Base Instruct

The best way Midm 2.0 Base Instruct runs on common GPUs, Macs and AI boxes
MachineBest fitContextSpeedFirst replyReport
RTX 4060Too bigPDF: Midm 2.0 Base Instruct on RTX 4060 report
RTX 5060Too bigPDF: Midm 2.0 Base Instruct on RTX 5060 report
RTX 3070Too bigPDF: Midm 2.0 Base Instruct on RTX 3070 report
RTX 3080 10 GB3-bit16K87 tok/s0.8 sPDF: Midm 2.0 Base Instruct on RTX 3080 10 GB report
RTX 3060 12 GB4-bit20K36 tok/s1.8 sPDF: Midm 2.0 Base Instruct on RTX 3060 12 GB report
Arc B5804-bit20K31 tok/s1.3 sPDF: Midm 2.0 Base Instruct on Arc B580 report
RTX 40704-bit20K49 tok/s0.8 sPDF: Midm 2.0 Base Instruct on RTX 4070 report
RTX 4070 Super4-bit20K49 tok/s0.7 sPDF: Midm 2.0 Base Instruct on RTX 4070 Super report
RTX 50704-bit20K64 tok/s0.8 sPDF: Midm 2.0 Base Instruct on RTX 5070 report
RTX 4060 Ti 16 GB6-bit28K22 tok/s1.1 sPDF: Midm 2.0 Base Instruct on RTX 4060 Ti 16 GB report
RTX 5060 Ti 16 GB6-bit28K33 tok/s1.0 sPDF: Midm 2.0 Base Instruct on RTX 5060 Ti 16 GB report
Arc A770 16 GB6-bit28K28 tok/s1.1 sPDF: Midm 2.0 Base Instruct on Arc A770 16 GB report
Radeon RX 9070 XT6-bit28K39 tok/s0.7 sPDF: Midm 2.0 Base Instruct on Radeon RX 9070 XT report
RTX 4070 Ti Super6-bit28K48 tok/s0.5 sPDF: Midm 2.0 Base Instruct on RTX 4070 Ti Super report
RTX 4080 Super6-bit28K53 tok/s0.4 sPDF: Midm 2.0 Base Instruct on RTX 4080 Super report
RTX 5070 Ti6-bit28K63 tok/s0.5 sPDF: Midm 2.0 Base Instruct on RTX 5070 Ti report
RTX 50806-bit28K67 tok/s0.4 sPDF: Midm 2.0 Base Instruct on RTX 5080 report
Radeon RX 7900 XTX8-bit32K44 tok/s1.0 sPDF: Midm 2.0 Base Instruct on Radeon RX 7900 XTX report
RTX 30908-bit32K52 tok/s0.7 sPDF: Midm 2.0 Base Instruct on RTX 3090 report
RTX 40908-bit32K56 tok/s0.3 sPDF: Midm 2.0 Base Instruct on RTX 4090 report
Radeon AI PRO R97008-bit32K30 tok/s0.7 sPDF: Midm 2.0 Base Instruct on Radeon AI PRO R9700 report
RTX 50908-bit32K93 tok/s0.2 sPDF: Midm 2.0 Base Instruct on RTX 5090 report
2× RTX 30908-bit32K52 tok/s0.4 sPDF: Midm 2.0 Base Instruct on 2× RTX 3090 report
2× RTX 40908-bit32K56 tok/s0.2 sPDF: Midm 2.0 Base Instruct on 2× RTX 4090 report
RTX PRO 5000 Blackwell 48 GB8-bit32K72 tok/s0.3 sPDF: Midm 2.0 Base Instruct on RTX PRO 5000 Blackwell 48 GB report
RTX 6000 Ada8-bit32K53 tok/s0.2 sPDF: Midm 2.0 Base Instruct on RTX 6000 Ada report
L40S8-bit32K48 tok/s0.4 sPDF: Midm 2.0 Base Instruct on L40S report
2× RTX 50908-bit32K93 tok/s0.1 sPDF: Midm 2.0 Base Instruct on 2× RTX 5090 report
4× RTX 30908-bit32K52 tok/s0.2 sPDF: Midm 2.0 Base Instruct on 4× RTX 3090 report
RTX PRO 6000 Blackwell8-bit32K93 tok/s0.2 sPDF: Midm 2.0 Base Instruct on RTX PRO 6000 Blackwell report
Mac mini M4 16 GB3-bit20K17 tok/s15 sPDF: Midm 2.0 Base Instruct on Mac mini M4 16 GB report
Mac mini M4 24 GB8-bit16K8.2 tok/s15 sPDF: Midm 2.0 Base Instruct on Mac mini M4 24 GB report
Mac mini M4 32 GB8-bit32K8.2 tok/s15 sPDF: Midm 2.0 Base Instruct on Mac mini M4 32 GB report
Mac mini M4 Pro 24 GB8-bit16K18 tok/s7.4 sPDF: Midm 2.0 Base Instruct on Mac mini M4 Pro 24 GB report
Mac mini M4 Pro 48 GB8-bit32K18 tok/s7.4 sPDF: Midm 2.0 Base Instruct on Mac mini M4 Pro 48 GB report
Mac mini M4 Pro 64 GB8-bit32K18 tok/s7.4 sPDF: Midm 2.0 Base Instruct on Mac mini M4 Pro 64 GB report
M4 Max 128 GB8-bit32K31 tok/s3.7 sPDF: Midm 2.0 Base Instruct on M4 Max 128 GB report
M5 Max 128 GB8-bit32K34 tok/s1.1 sPDF: Midm 2.0 Base Instruct on M5 Max 128 GB report
M3 Ultra 96 GB8-bit32K39 tok/s2.1 sPDF: Midm 2.0 Base Instruct on M3 Ultra 96 GB report
M3 Ultra 256 GB8-bit32K39 tok/s2.1 sPDF: Midm 2.0 Base Instruct on M3 Ultra 256 GB report
M3 Ultra 512 GB8-bit32K39 tok/s2.1 sPDF: Midm 2.0 Base Instruct on M3 Ultra 512 GB report
M5 Ultra 512 GB8-bit32K53 tok/s0.7 sPDF: Midm 2.0 Base Instruct on M5 Ultra 512 GB report
Ryzen AI Max+ 395 64 GB8-bit32K13 tok/s2.1 sPDF: Midm 2.0 Base Instruct on Ryzen AI Max+ 395 64 GB report
Ryzen AI Max+ 395 128 GB8-bit32K13 tok/s2.1 sPDF: Midm 2.0 Base Instruct on Ryzen AI Max+ 395 128 GB report
RTX Spark8-bit32K18 tok/s0.8 sPDF: Midm 2.0 Base Instruct on RTX Spark report
DGX Spark8-bit32K16 tok/s0.8 sPDF: Midm 2.0 Base Instruct on DGX Spark report
Dual DGX Spark8-bit32K16 tok/s0.4 sPDF: Midm 2.0 Base Instruct on Dual DGX Spark report
DGX Station GB3008-bit32K264 tok/s0.1 sPDF: Midm 2.0 Base Instruct on DGX Station GB300 report
2× RTX PRO 6000 Blackwell8-bit32K93 tok/s0.1 sPDF: Midm 2.0 Base Instruct on 2× RTX PRO 6000 Blackwell report
4× RTX PRO 6000 Blackwell8-bit32K93 tok/s0.1 sPDF: Midm 2.0 Base Instruct on 4× RTX PRO 6000 Blackwell report
8× RTX PRO 6000 Blackwell8-bit32K93 tok/s0.1 sPDF: Midm 2.0 Base Instruct on 8× RTX PRO 6000 Blackwell report
A100 80 GB SXM8-bit32K104 tok/s0.5 sPDF: Midm 2.0 Base Instruct on A100 80 GB SXM report
H100 SXM8-bit32K156 tok/s0.2 sPDF: Midm 2.0 Base Instruct on H100 SXM report
4× H200 NVL8-bit32K203 tok/s0.1 sPDF: Midm 2.0 Base Instruct on 4× H200 NVL report
8× H200 NVL8-bit32K203 tok/s0.1 sPDF: Midm 2.0 Base Instruct 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 gchrisoh/Midm-2.0-Base-Instruct-Q4_K_M-GGUF
vLLM
vllm serve K-intelligence/Midm-2.0-Base-Instruct

Questions

How much VRAM does Midm 2.0 Base Instruct need?

Midm 2.0 Base Instruct needs 6.5 GB for its weights at 4-bit, 8.8 GB with an 8K context and 13.3 GB with 32K. At 8-bit, 13.8 GB.

What is the minimum VRAM for Midm 2.0 Base Instruct?

About 7.6 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 Midm 2.0 Base Instruct?

11.5B parameters: 21.5 GB of weights at 16-bit, 6.5 GB at 4-bit.

Can Midm 2.0 Base Instruct run on an RTX 4090?

Yes. At 8-bit it holds 32K of context and writes about 56 tokens a second.

More models