How much VRAM does OpenCoder-1.5B-Instruct need?

Q4_K_M is exactly 1,417,775,200 bytes (1.32 GiB) in weights. Add the KV cache, which depends on context length, plus roughly half a gigabyte of runtime overhead.

How large is OpenCoder-1.5B-Instruct's KV cache?

6.56 GiB at 32K context with an f16 cache, computed per layer from the model's architecture. Quantizing the cache to q8_0 roughly halves it.

infly · text

OpenCoder-1.5B-Instruct

infly/OpenCoder-1.5B-Instruct

OpenCoder-1.5B-Instruct at Q4_K_M is exactly 1,417,775,200 bytes (1.32 GiB / 1.42 GB) — an effective 5.952 bits per weight, not the nominal 4. Its KV cache at 32K is 6.56 GiB.

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

Parameters

1.9B

Architecture

llama

24 layers

Context

4,096

native (config.json)

License

other

Shipped quantizations

● exact bytes, summed from published files

Quant	Size●	Exact bytes●	Effective bpw●	Tensors●	Publisher
Q3_K_L	1.18 GiB	1,269,218,400	5.328	—	lmstudio-community
Q4_K_M	1.32 GiB	1,417,775,200	5.952	—	lmstudio-community
Q6_K	1.81 GiB	1,947,311,200	8.175	—	lmstudio-community
Q8_0	1.89 GiB	2,027,306,080	8.511	—	lmstudio-community

KV cache by context

computed per layer

Context	KV cache (f16)●	Flat formula	Overstated by	Full / windowed / recurrent
4,096	0.82 GiB	0.82 GiB	—	24 / 0 / 0
8,192	1.64 GiB	1.64 GiB	—	24 / 0 / 0
16,384	3.28 GiB	3.28 GiB	—	24 / 0 / 0
32,768	6.56 GiB	6.56 GiB	—	24 / 0 / 0
65,536	13.13 GiB	13.13 GiB	—	24 / 0 / 0
131,072	26.25 GiB	26.25 GiB	—	24 / 0 / 0

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 1.00 GiB. The real file is 1.32 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

160

Hidden size

2240

Vocab

96,640

Sliding window

none

SWA period

—

MLA

Experts

—

Experts per token

—

use_sliding_window

—

Questions people ask

How much VRAM does OpenCoder-1.5B-Instruct need?: Q4_K_M is exactly 1,417,775,200 bytes (1.32 GiB) in weights. Add the KV cache, which depends on your context length, plus roughly half a gigabyte of runtime overhead.
How large is OpenCoder-1.5B-Instruct's KV cache?: 6.56 GiB at 32K context with an f16 cache, computed per layer. Quantizing the cache to q8_0 roughly halves it, which is often the difference between a context length fitting and not.
Which quantization of OpenCoder-1.5B-Instruct 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.