Skip to content

Best GPU for Qwen2.5 Coder 32B: VRAM Requirements

Qwen2.5 Coder 32B needs about 73.9 GB of GPU memory in BF16 and 20.2 GB at 4 bits. See which GPUs fit it for inference and fine-tuning.

Qwen2.5 Coder 32B needs about 73.9 GB of VRAM in FP16, 38.1 GB in FP8 and 20.2 GB at 4 bits for inference with an 8,192-token context. In FP16 it fits on one NVIDIA RTX PRO 6000 Blackwell 96GB; a 4-bit build fits on one NVIDIA GeForce RTX 4090 24GB.

Qwen2.5 Coder 32B is a 32.5B-parameter language model from Alibaba Cloud (Qwen team), released in November 2024, with a 32,768-token native context window that extends to 131,072 tokens with YaRN and published under the Apache 2.0. Its weights alone take 65 GB in BF16 and 16.3 GB at 4 bits per parameter; the tables below add the KV cache and runtime memory, then list the GPUs that hold each setup.

Recommended GPUs for Qwen2.5 Coder 32B

Qwen2.5 Coder 32B VRAM requirements by precision

One request with an 8,192-token context, KV cache in BF16, plus a 10 % margin for activations and runtime buffers.

PrecisionWeightsKV cacheTotal for inference
BF16 / FP1665 GB2.1 GB73.9 GB
FP832.5 GB2.1 GB38.1 GB
INT416.3 GB2.1 GB20.2 GB

Qwen2.5 Coder 32B VRAM by task

Fine-tuning rows assume a short sequence, batch 1 and gradient checkpointing; longer sequences add activation memory.

TaskGPU memory
Inference at the full 131,072-token context (FP8)73.5 GB
QLoRA (4-bit base)21.1 GB
LoRA (BF16 base)84.5 GB
Full fine-tuning (BF16, AdamW)572 GB

GPUs that fit Qwen2.5 Coder 32B

Number of cards whose combined memory holds each setup, assuming the model is split across them with tensor or pipeline parallelism. A dash means more than eight cards.

GPUMemoryBF16 inference (73.9 GB)FP8 inference (38.1 GB)INT4 inference (20.2 GB)QLoRA (21.1 GB)LoRA (84.5 GB)
NVIDIA GeForce RTX 5080 16GB16 GB8 GPUs4 GPUs2 GPUs2 GPUs8 GPUs
NVIDIA GeForce RTX 4090 24GB24 GB4 GPUs2 GPUs1 GPU1 GPU4 GPUs
NVIDIA L4 24GB24 GB4 GPUs2 GPUs1 GPU1 GPU4 GPUs
NVIDIA GeForce RTX 5090 32GB32 GB4 GPUs2 GPUs1 GPU1 GPU4 GPUs
NVIDIA Tesla V100 32GB32 GB4 GPUs2 GPUs1 GPU1 GPU4 GPUs
NVIDIA A100 PCIe 40GB40 GB2 GPUs1 GPU1 GPU1 GPU4 GPUs
NVIDIA L40S 48GB48 GB2 GPUs1 GPU1 GPU1 GPU2 GPUs
NVIDIA Quadro RTX 8000 48GB48 GB2 GPUs1 GPU1 GPU1 GPU2 GPUs
NVIDIA RTX 6000 Ada 48GB48 GB2 GPUs1 GPU1 GPU1 GPU2 GPUs
NVIDIA RTX A6000 48GB48 GB2 GPUs1 GPU1 GPU1 GPU2 GPUs
NVIDIA A100 PCIe 80GB80 GB1 GPU1 GPU1 GPU1 GPU2 GPUs
NVIDIA A100 SXM 80GB80 GB1 GPU1 GPU1 GPU1 GPU2 GPUs
NVIDIA H100 PCIe 80GB80 GB1 GPU1 GPU1 GPU1 GPU2 GPUs
NVIDIA H100 SXM 80GB80 GB1 GPU1 GPU1 GPU1 GPU2 GPUs
NVIDIA RTX PRO 6000 Blackwell 96GB96 GB1 GPU1 GPU1 GPU1 GPU1 GPU
NVIDIA H200 141GB141 GB1 GPU1 GPU1 GPU1 GPU1 GPU
NVIDIA B200 180GB180 GB1 GPU1 GPU1 GPU1 GPU1 GPU
NVIDIA B300 288GB288 GB1 GPU1 GPU1 GPU1 GPU1 GPU
NVIDIA GB300 NVL288 GB1 GPU1 GPU1 GPU1 GPU1 GPU

What is Qwen2.5 Coder 32B?

Qwen2.5 Coder 32B is a code-specialized model trained on 5.5 trillion tokens of source code, text-code grounding and synthetic data. It is used for code generation, code repair and repository-level completion, often served next to an editor or an agent.

Developer
Alibaba Cloud (Qwen team)
Released
November 2024
Parameters
32.5B
Context window
32,768 tokens native, 131,072 with YaRN
License
Apache 2.0

The model card supports 131,072 tokens with YaRN scaling; the native window is 32,768 tokens.

Source: Qwen2.5 Coder 32B Instruct model card (as of November 17, 2026).

Frequently asked questions

How much VRAM does Qwen2.5 Coder 32B need?
For inference with an 8,192-token context and one request, Qwen2.5 Coder 32B needs about 73.9 GB in BF16, 38.1 GB in FP8, 20.2 GB in INT4, including a 10 % runtime margin. Real usage depends on the inference engine and its settings.
Can Qwen2.5 Coder 32B run on a single GPU?
Yes. In BF16 it needs about 73.9 GB, which fits one NVIDIA RTX PRO 6000 Blackwell 96GB (96 GB).
How much GPU memory does fine-tuning Qwen2.5 Coder 32B take?
QLoRA (4-bit base): about 21.1 GB; LoRA (BF16 base): about 84.5 GB; Full fine-tuning (BF16, AdamW): about 572 GB. Longer training sequences and larger batches add activation memory on top.
What is the best GPU for Qwen2.5 Coder 32B?
Best for inference: 1 GPU × NVIDIA L40S 48GB (FP8 or INT8 weights, 38.1 GB at 8,192 tokens). Best for fine-tuning: 1 GPU × NVIDIA H200 141GB (LoRA on BF16 weights, 84.5 GB). Budget pick: 1 GPU × NVIDIA GeForce RTX 4090 24GB (Smallest card class that holds a 4-bit build, 20.2 GB). Check live offers before booking, as availability changes.
Can Qwen2.5 Coder 32B run on an RTX 4090?
Yes. In 4-bit it needs about 20.2 GB, which fits one RTX 4090 (24 GB). FP16 needs about 73.9 GB, 4 of these cards.
Can Qwen2.5 Coder 32B run on an RTX 5090?
Yes. In 4-bit it needs about 20.2 GB, which fits one RTX 5090 (32 GB). FP16 needs about 73.9 GB, 4 of these cards.

Estimates use the formulas of the LLM VRAM calculator (decimal GB). Real usage depends on the inference engine, quantization format and memory fragmentation: test on the target GPU before a long rental.

Explore the live market

Every offer on the live market comes from a reviewed supplier and is sold by EverXCloud Inc.