LimitDigest
Home / Models / Local finder / Requirements / Inkling 2 Mini

Inkling 2 Mini VRAM requirements

Thinking Machines · 26.4B params (5.4B active) · Apache 2.0. Estimates, not vendor numbers.

Quick answer

Inkling 2 Mini (26.4B parameters, 5.4B active per token) needs about 17 GB of VRAM or unified memory at Q4, 20 GB at Q5 and 30 GB at Q8 (weights plus about 1.5 GB headroom, our estimate). The smallest common setup that runs it is 24GB GPU (RTX 4090).

QuantDownloadMemory needed
Q416 GB17 GB
Q519 GB20 GB
Q828 GB30 GB

Will it run on my machine?

SetupRuns?Speed est.
8GB GPU (RTX 4060)Non/a
12GB GPU (RTX 3060 / 4070)Non/a
16GB GPU (RTX 4060 Ti 16GB)Non/a
24GB GPU (RTX 4090)Yes, Q5~157 tok/s
32GB GPU (RTX 5090)Yes, Q8~187 tok/s
Mac 16GB (M4)Non/a
Mac 48GB (M4 Pro)Yes, Q8~29 tok/s
Mac 128GB (M4 Max)Yes, Q8~57 tok/s
PC, 32GB RAM, no GPUYes, Q4~11 tok/s

Best quant that fits each setup, with estimated generation speed from memory bandwidth. Real speed depends on the runtime, context length and settings.

Looped mixture of experts: 26.4B total, 3.4B active, and the middle 12 layers run three times, so about 5.4B weights are read per token. 1M context, text and image input. Open weights released Nov 20, 2026. The official checkpoint (NVFP4 plus MXFP8) is about 20 GB; Thinking Machines recommends a Mac with 32 GB or more, or 32 GB of VRAM.

FAQ

How much VRAM does Inkling 2 Mini need?

About 17 GB at Q4, 20 GB at Q5 and 30 GB at Q8, including about 1.5 GB headroom. These are estimates from the parameter count on the official model card.

Can Inkling 2 Mini run on 8GB of VRAM?

Not fully on the GPU. At Q4 it needs about 17 GB; partial CPU offload works but is much slower.

Can Inkling 2 Mini run on a Mac?

Yes. Mac 48GB (M4 Pro), Mac 128GB (M4 Max) can run it.

Which quant of Inkling 2 Mini should I use?

Use the highest quant that fits with room to spare: Q8 is near full quality, Q5 is a good balance, and Q4 is the smallest that usually keeps quality acceptable.

Why is Inkling 2 Mini fast for its size?

It is a mixture of experts: 26.4B parameters in total but only 5.4B are used per token, so it reads less memory per token. You still need memory for all 26.4B.

Similar-size models

LFM2.5 24B A2B (24B) · Gemma 4 26B A4B (25.2B) · Qwen3.6 27B (27B) · Qwen3.8 27B (27.8B) · Muse Glimmer 30B (29.6B) · Qwen3 Coder 30B A3B (30.5B)