How much VRAM does dolphin-2_6-phi-2 need?

Q4_K_M is exactly 1,789,240,608 bytes (1.67 GiB) in weights. Add the KV cache, which depends on context length, plus roughly half a gigabyte of runtime overhead.

dphn · text

dolphin-2_6-phi-2

dphn/dolphin-2_6-phi-2

dolphin-2_6-phi-2 at Q4_K_M is exactly 1,789,240,608 bytes (1.67 GiB / 1.79 GB) — an effective 5.143 bits per weight, not the nominal 4.

From the file· summed from 1 file(s)From the file· KV per layer

Parameters

2.8B

Architecture

phi2

32 layers

Context

—

native (config.json)

License

other

Shipped quantizations

● exact bytes, summed from published files

Quant	Size●	Exact bytes●	Effective bpw●	Tensors●	Publisher
Q2_K	1.09 GiB	1,173,611,808	3.373	—	TheBloke
Q3_K_S	1.16 GiB	1,250,821,408	3.595	—	TheBloke
Q3_K_M	1.38 GiB	1,480,197,408	4.255	—	TheBloke
Q4_0	1.49 GiB	1,602,463,008	4.606	—	TheBloke
Q3_K_L	1.49 GiB	1,604,715,808	4.613	—	TheBloke
Q4_K_S	1.50 GiB	1,615,570,208	4.644	—	TheBloke
Q4_K_M	1.67 GiB	1,789,240,608	5.143	—	TheBloke
Q5_K_S	1.80 GiB	1,933,419,808	5.558	—	TheBloke
Q5_0	1.80 GiB	1,933,419,808	5.558	—	TheBloke
Q5_K_M	1.93 GiB	2,072,683,808	5.958	—	TheBloke
Q6_K	2.13 GiB	2,285,061,408	6.568	—	TheBloke
Q8_0	2.75 GiB	2,958,034,208	8.503	—	TheBloke

Compare with

same modality, comparable size

vs Qwen3-4B vs Qwen3-8B vs HyperCLOVAX-SEED-Text-Instruct-1.5B vs Llama-3.2-1B-Instruct vs Qwen3-VL-8B-Instruct-abliterated-v1 vs Llama-3.1-8B-Instruct All alternatives →Will it fit on my card? →

Why other calculators give a different number

A parameters × bits ÷ 8 estimate puts Q4_K_M at roughly 1.46 GiB. The real file is 1.67 GiB, because a quantization is a mixture and some tensors are always kept at higher precision.

Architecture

from config.json

Layers

Attention heads

—

KV heads

—

Head dim

—

Hidden size

—

Vocab

51,200

Sliding window

none

SWA period

—

MLA

Experts

—

Experts per token

—

use_sliding_window

—

Questions people ask

How much VRAM does dolphin-2_6-phi-2 need?: Q4_K_M is exactly 1,789,240,608 bytes (1.67 GiB) in weights. Add the KV cache, which depends on your context length, plus roughly half a gigabyte of runtime overhead.
Which quantization of dolphin-2_6-phi-2 should I use?: Q4_K_M is the usual default. Pick the largest quantization that fits your card at the context you actually need — the table above gives exact sizes for every one published.