Best GPU for Qwen3 8B: VRAM Requirements
Qwen3 8B needs about 19.4 GB of GPU memory in BF16 and 5.8 GB at 4 bits. See which GPUs fit it for inference and fine-tuning.
Qwen3 8B needs about 19.4 GB of VRAM in FP16, 10.3 GB in FP8 and 5.8 GB at 4 bits for inference with an 8,192-token context. In FP16 it fits on one NVIDIA GeForce RTX 4090 24GB; a 4-bit build fits on one NVIDIA GeForce RTX 5080 16GB.
Qwen3 8B is a 8.2B-parameter language model from Alibaba Cloud (Qwen team), released in April 2025, 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 16.4 GB in BF16 and 4.1 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 Qwen3 8B
Best for inference
1 GPU × NVIDIA L4 24GB
FP8 or INT8 weights, 10.3 GB at 8,192 tokens.
Best for fine-tuning
1 GPU × NVIDIA L4 24GB
LoRA on BF16 weights, 21.3 GB.
Budget pick
1 GPU × NVIDIA GeForce RTX 5080 16GB
Smallest card class that holds a 4-bit build, 5.8 GB.
Qwen3 8B 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.
| Precision | Weights | KV cache | Total for inference |
|---|---|---|---|
| BF16 / FP16 | 16.4 GB | 1.2 GB | 19.4 GB |
| FP8 | 8.2 GB | 1.2 GB | 10.3 GB |
| INT4 | 4.1 GB | 1.2 GB | 5.8 GB |
Qwen3 8B VRAM by task
Fine-tuning rows assume a short sequence, batch 1 and gradient checkpointing; longer sequences add activation memory.
| Task | GPU memory |
|---|---|
| Inference at the full 131,072-token context (FP8) | 30.3 GB |
| QLoRA (4-bit base) | 5.3 GB |
| LoRA (BF16 base) | 21.3 GB |
| Full fine-tuning (BF16, AdamW) | 144.3 GB |
GPUs that fit Qwen3 8B
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.
| GPU | Memory | BF16 inference (19.4 GB) | FP8 inference (10.3 GB) | INT4 inference (5.8 GB) | QLoRA (5.3 GB) | LoRA (21.3 GB) |
|---|---|---|---|---|---|---|
| NVIDIA GeForce RTX 5080 16GB | 16 GB | 2 GPUs | 1 GPU | 1 GPU | 1 GPU | 2 GPUs |
| NVIDIA GeForce RTX 4090 24GB | 24 GB | 1 GPU | 1 GPU | 1 GPU | 1 GPU | 1 GPU |
| NVIDIA L4 24GB | 24 GB | 1 GPU | 1 GPU | 1 GPU | 1 GPU | 1 GPU |
| NVIDIA GeForce RTX 5090 32GB | 32 GB | 1 GPU | 1 GPU | 1 GPU | 1 GPU | 1 GPU |
| NVIDIA Tesla V100 32GB | 32 GB | 1 GPU | 1 GPU | 1 GPU | 1 GPU | 1 GPU |
| NVIDIA A100 PCIe 40GB | 40 GB | 1 GPU | 1 GPU | 1 GPU | 1 GPU | 1 GPU |
| NVIDIA L40S 48GB | 48 GB | 1 GPU | 1 GPU | 1 GPU | 1 GPU | 1 GPU |
| NVIDIA Quadro RTX 8000 48GB | 48 GB | 1 GPU | 1 GPU | 1 GPU | 1 GPU | 1 GPU |
| NVIDIA RTX 6000 Ada 48GB | 48 GB | 1 GPU | 1 GPU | 1 GPU | 1 GPU | 1 GPU |
| NVIDIA RTX A6000 48GB | 48 GB | 1 GPU | 1 GPU | 1 GPU | 1 GPU | 1 GPU |
| NVIDIA A100 PCIe 80GB | 80 GB | 1 GPU | 1 GPU | 1 GPU | 1 GPU | 1 GPU |
| NVIDIA A100 SXM 80GB | 80 GB | 1 GPU | 1 GPU | 1 GPU | 1 GPU | 1 GPU |
| NVIDIA H100 PCIe 80GB | 80 GB | 1 GPU | 1 GPU | 1 GPU | 1 GPU | 1 GPU |
| NVIDIA H100 SXM 80GB | 80 GB | 1 GPU | 1 GPU | 1 GPU | 1 GPU | 1 GPU |
| NVIDIA RTX PRO 6000 Blackwell 96GB | 96 GB | 1 GPU | 1 GPU | 1 GPU | 1 GPU | 1 GPU |
| NVIDIA H200 141GB | 141 GB | 1 GPU | 1 GPU | 1 GPU | 1 GPU | 1 GPU |
| NVIDIA B200 180GB | 180 GB | 1 GPU | 1 GPU | 1 GPU | 1 GPU | 1 GPU |
| NVIDIA B300 288GB | 288 GB | 1 GPU | 1 GPU | 1 GPU | 1 GPU | 1 GPU |
| NVIDIA GB300 NVL | 288 GB | 1 GPU | 1 GPU | 1 GPU | 1 GPU | 1 GPU |
What is Qwen3 8B?
Qwen3 8B is a dense model of the Qwen3 generation that switches between a thinking mode for step-by-step reasoning and a direct mode for fast replies, in one set of weights. It supports more than 100 languages and dialects.
- Developer
- Alibaba Cloud (Qwen team)
- Released
- April 2025
- Parameters
- 8.2B
- 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: Qwen3 8B model card (as of November 17, 2026).
Frequently asked questions
- How much VRAM does Qwen3 8B need?
- For inference with an 8,192-token context and one request, Qwen3 8B needs about 19.4 GB in BF16, 10.3 GB in FP8, 5.8 GB in INT4, including a 10 % runtime margin. Real usage depends on the inference engine and its settings.
- Can Qwen3 8B run on a single GPU?
- Yes. In BF16 it needs about 19.4 GB, which fits one NVIDIA GeForce RTX 4090 24GB (24 GB).
- How much GPU memory does fine-tuning Qwen3 8B take?
- QLoRA (4-bit base): about 5.3 GB; LoRA (BF16 base): about 21.3 GB; Full fine-tuning (BF16, AdamW): about 144.3 GB. Longer training sequences and larger batches add activation memory on top.
- What is the best GPU for Qwen3 8B?
- Best for inference: 1 GPU × NVIDIA L4 24GB (FP8 or INT8 weights, 10.3 GB at 8,192 tokens). Best for fine-tuning: 1 GPU × NVIDIA L4 24GB (LoRA on BF16 weights, 21.3 GB). Budget pick: 1 GPU × NVIDIA GeForce RTX 5080 16GB (Smallest card class that holds a 4-bit build, 5.8 GB). Check live offers before booking, as availability changes.
- Can Qwen3 8B run on an RTX 4090?
- Yes. In FP16 it needs about 19.4 GB, which fits one RTX 4090 (24 GB).
- Can Qwen3 8B run on an RTX 5090?
- Yes. In FP16 it needs about 19.4 GB, which fits one RTX 5090 (32 GB).
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.