Magic Tools
LLM VRAM Calculator (all models)

Can You Run Qwen3-Coder-30B-A3B? VRAM Requirements

Qwen3-Coder-30B-A3B has 30.5B total parameters (MoE, 3.3B active per token). Running it at Q4_K_M (GGUF) with 16K context takes roughly 20.3 GB of memory (17.3 GB weights + 1.5 GB KV cache + runtime overhead). The minimum viable hardware is about RTX 3090 / 4090 24GB. All numbers are planning estimates, verified 2026-09-05.

VRAM by quantization (16K context)

QuantizationWeightsTotal neededMinimum hardware
FP16 / BF1656.8 GB62.9 GBA100 / H100 80GB
FP828.4 GB32.2 GBMac 36GB unified
Q8_0 (GGUF)30.1 GB34.0 GBA100 40GB
Q6_K (GGUF)23.3 GB26.7 GBRTX 5090 32GB
Q5_K_M (GGUF)20.2 GB23.3 GBRTX 5090 32GB
MXFP415.1 GB18.1 GBRTX 3090 / 4090 24GB
NVFP415.1 GB18.1 GBRTX 3090 / 4090 24GB
Q4_K_M (GGUF)recommended17.3 GB20.3 GBRTX 3090 / 4090 24GB
Q3_K_M (GGUF)13.4 GB16.4 GBRTX 3090 / 4090 24GB
Q2_K (GGUF)9.9 GB12.9 GBRTX 4060 Ti 16GB

How context length changes memory (Q4_K_M (GGUF))

ContextKV cacheTotal needed
4K0.4 GB19.2 GB
16K1.5 GB20.3 GB
64K6.0 GB24.8 GB
128K12.0 GB30.8 GB
256K24.0 GB42.8 GB

KV cache grows linearly with context; FP8 KV cache halves it again (try it in the calculator below).

Customize the estimate

Back

LLM VRAM Calculator — Can I Run It?

Estimate how much GPU memory (VRAM) you need to run open-weight LLMs like Kimi K3, DeepSeek R1, Qwen3, or Llama locally. Pick a model, quantization, and context length — the calculator adds up weights, KV cache, and runtime overhead, then shows which GPUs or Macs can fit it. Runs entirely in your browser: no upload, no signup.

30.5B total params · 3.3B active (MoE) · max 256K context

Recommended balance · ~0.61 bytes/param

Estimated memory needed

30.8 GB

Model weights17.3 GB
KV cache (131,072 tokens)12.0 GB
Runtime overhead1.5 GB

Will it fit?

RTX 3060 12GB3× needed
RTX 4060 Ti 16GB3× needed
RTX 3090 / 4090 24GB2× needed
RTX 5090 32GB2× needed
A100 40GB✓ fits
A100 / H100 80GB✓ fits
H200 141GB✓ fits
AMD MI300X 192GB✓ fits
B200 192GB✓ fits
Mac 24GB unified✗ too big
Mac 36GB unified✓ fits
Mac 64GB unified✓ fits
Mac 128GB unified✓ fits
Mac M3 Ultra 512GB✓ fits

Assumes ~92% of device memory is usable. Multi-GPU counts are for tensor/pipeline parallel serving (vLLM, SGLang); Apple Silicon uses unified memory via llama.cpp or MLX.

FAQ

How much VRAM does Qwen3-Coder-30B-A3B need?

About 20.3 GB at the recommended Q4_K_M (GGUF) quantization with 16K context: 17.3 GB for weights, 1.5 GB for KV cache, plus runtime overhead. Longer context grows the KV cache.

What is the minimum hardware for Qwen3-Coder-30B-A3B?

Roughly RTX 3090 / 4090 24GB, assuming 92% of device memory is usable. Lower-end setups can try more aggressive quantization (Q3/Q2) at a noticeable quality cost.

Can I run Qwen3-Coder-30B-A3B on a Mac?

Yes. At Q4_K_M (GGUF) with 16K context it needs about 20.3 GB, so a Mac 24GB unified or larger fits it via llama.cpp or MLX. Note macOS only lets the GPU wire roughly 2/3-3/4 of RAM by default; near the limit, raise iogpu.wired_limit_mb or use a smaller quant.

Which quantization should I use for Qwen3-Coder-30B-A3B?

Q4_K_M is the recommended balance of size and quality; with headroom, Q6_K or Q8_0 reduce quality loss further.

VRAM requirements for other models