How much VRAM does Yi-Coder-1.5B need?

Q4_K_M is exactly 963,673,472 bytes (0.90 GiB) in weights. Add the KV cache, which depends on context length, plus roughly half a gigabyte of runtime overhead.

How large is Yi-Coder-1.5B's KV cache?

6.00 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.

01-ai · text

Yi-Coder-1.5B

01-ai/Yi-Coder-1.5B

Yi-Coder-1.5B at Q4_K_M is exactly 963,673,472 bytes (0.90 GiB / 0.96 GB) — an effective 5.221 bits per weight, not the nominal 4. Its KV cache at 32K is 6.00 GiB.

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

Parameters

1.5B

Architecture

llama

24 layers

Context

131,072

native (config.json)

License

apache-2.0

Shipped quantizations

● exact bytes, summed from published files

Quant	Size●	Exact bytes●	Effective bpw●	Tensors●	Publisher
IQ1_S	0.46 GiB	491,167,360	2.661	—	legraphista
IQ1_M	0.47 GiB	508,567,168	2.756	—	legraphista
IQ2_XXS	0.50 GiB	537,566,848	2.913	—	legraphista
IQ2_XS	0.53 GiB	563,912,320	3.055	—	legraphista
IQ2_S	0.56 GiB	601,603,712	3.260	—	legraphista
Q2_K_S	0.58 GiB	618,479,232	3.351	—	legraphista
IQ2_M	0.58 GiB	624,803,456	3.385	—	legraphista
Q2_K	0.59 GiB	634,699,392	3.439	—	legraphista
IQ3_XXS	0.61 GiB	653,901,440	3.543	—	legraphista
IQ3_XS	0.65 GiB	694,951,552	3.765	—	legraphista
IQ3_S	0.67 GiB	723,410,560	3.920	—	legraphista
Q3_K_S	0.67 GiB	723,410,560	3.920	—	legraphista
IQ3_M	0.70 GiB	754,376,320	4.087	—	legraphista
IQ3_M	0.70 GiB	754,376,320	4.087	—	bartowski
Q3_K	0.73 GiB	785,718,912	4.257	—	legraphista
Q3_K_L	0.77 GiB	826,039,936	4.476	—	legraphista
Q3_K_L	0.77 GiB	826,039,936	4.476	—	bartowski
IQ4_XS	0.78 GiB	832,568,960	4.511	—	legraphista
IQ4_XS	0.78 GiB	832,568,960	4.511	—	bartowski
IQ4_NL	0.81 GiB	866,156,160	4.693	—	legraphista
Q4_0	0.81 GiB	868,269,696	4.705	—	bartowski
Q4_K_S	0.84 GiB	904,183,424	4.899	—	legraphista
Q4_K_S	0.84 GiB	904,183,424	4.899	—	bartowski
Q4_K_M	0.90 GiB	963,673,472	5.221	—	lmstudio-community
Q4_K_M	0.90 GiB	963,673,728	5.221	—	bartowski
Q4_K	0.90 GiB	963,673,728	5.221	—	legraphista
Q5_K_S	0.98 GiB	1,051,229,568	5.696	—	legraphista
Q5_K_S	0.98 GiB	1,051,229,824	5.696	—	bartowski
Q4_K_L	0.99 GiB	1,060,953,728	5.748	—	bartowski
Q5_K	1.02 GiB	1,100,184,960	5.961	—	legraphista
Q5_K_M	1.02 GiB	1,100,185,216	5.961	—	bartowski
Q5_K_L	1.10 GiB	1,181,081,216	6.399	—	bartowski
Q6_K	1.19 GiB	1,278,516,608	6.927	—	legraphista
Q6_K	1.19 GiB	1,278,516,608	6.927	—	lmstudio-community
Q6_K	1.19 GiB	1,278,516,864	6.927	—	bartowski
Q6_K_L	1.25 GiB	1,342,004,864	7.271	—	bartowski
Q8_0	1.46 GiB	1,570,561,408	8.510	—	lmstudio-community
Q8_0	1.46 GiB	1,570,561,408	8.510	—	legraphista
Q8_0	1.46 GiB	1,570,561,664	8.510	—	bartowski
F16	2.75 GiB	2,954,681,728	16.009	—	bartowski

KV cache by context

computed per layer

Context	KV cache (f16)●	Flat formula	Overstated by	Full / windowed / recurrent
4,096	0.75 GiB	0.75 GiB	—	24 / 0 / 0
8,192	1.50 GiB	1.50 GiB	—	24 / 0 / 0
16,384	3.00 GiB	3.00 GiB	—	24 / 0 / 0
32,768	6.00 GiB	6.00 GiB	—	24 / 0 / 0
65,536	12.00 GiB	12.00 GiB	—	24 / 0 / 0
131,072	24.00 GiB	24.00 GiB	—	24 / 0 / 0

Compare with

same modality, comparable size

vs Qwen3-4B vs HyperCLOVAX-SEED-Text-Instruct-1.5B vs Llama-3.2-1B-Instruct vs ced-base vs gemma-3-1b-it vs Qwen3-1.7B 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 0.77 GiB. The real file is 0.90 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

128

Hidden size

2048

Vocab

64,000

Sliding window

none

SWA period

—

MLA

Experts

—

Experts per token

—

use_sliding_window

—

Questions people ask

How much VRAM does Yi-Coder-1.5B need?: Q4_K_M is exactly 963,673,472 bytes (0.90 GiB) in weights. Add the KV cache, which depends on your context length, plus roughly half a gigabyte of runtime overhead.
How large is Yi-Coder-1.5B's KV cache?: 6.00 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 Yi-Coder-1.5B 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.