Best GPU for Whisper large-v3: VRAM Requirements
Whisper large-v3 needs about 3.4 GB of GPU memory in BF16 and 0.9 GB at 4 bits. See which GPUs fit it for inference and fine-tuning.
Whisper large-v3 needs about 3.4 GB of VRAM in FP16, 1.7 GB in FP8 and 0.9 GB at 4 bits for inference. In FP16 it fits on one NVIDIA GeForce RTX 5080 16GB.
Whisper large-v3 is a 1.6B-parameter speech recognition model from OpenAI, released in November 2023 and published under the MIT. Its weights alone take 3.1 GB in BF16 and 0.8 GB at 4 bits per parameter; the tables below add runtime memory and fine-tuning, then list the GPUs that hold each setup.
Recommended GPUs for Whisper large-v3
Best for inference
1 GPU × NVIDIA L4 24GB
FP8 or INT8 weights, 1.7 GB.
Best for fine-tuning
1 GPU × NVIDIA L4 24GB
LoRA on BF16 weights, 4 GB.
Budget pick
1 GPU × NVIDIA GeForce RTX 5080 16GB
Smallest card class that holds a 4-bit build, 0.9 GB.
Whisper large-v3 VRAM requirements by precision
Model weights plus a 10 % margin for activations and runtime buffers, at one request.
| Precision | Weights | KV cache | Total for inference |
|---|---|---|---|
| BF16 / FP16 | 3.1 GB | Not applicable | 3.4 GB |
| FP8 | 1.6 GB | Not applicable | 1.7 GB |
| INT4 | 0.8 GB | Not applicable | 0.9 GB |
Whisper large-v3 VRAM by task
Fine-tuning rows assume a short sequence, batch 1 and gradient checkpointing; longer sequences add activation memory.
| Task | GPU memory |
|---|---|
| QLoRA (4-bit base) | 1 GB |
| LoRA (BF16 base) | 4 GB |
| Full fine-tuning (BF16, AdamW) | 27.3 GB |
GPUs that fit Whisper large-v3
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 (3.4 GB) | FP8 inference (1.7 GB) | INT4 inference (0.9 GB) | QLoRA (1 GB) | LoRA (4 GB) |
|---|---|---|---|---|---|---|
| NVIDIA GeForce RTX 5080 16GB | 16 GB | 1 GPU | 1 GPU | 1 GPU | 1 GPU | 1 GPU |
| 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 Whisper large-v3?
Whisper large-v3 is OpenAI's largest speech recognition model, an encoder-decoder transformer trained on about 5 million hours of audio for transcription and translation into English in close to 100 languages. It processes audio in 30-second windows.
- Developer
- OpenAI
- Released
- November 2023
- Parameters
- 1.6B
- License
- MIT
Source: Whisper large-v3 model card (as of November 17, 2026).
Frequently asked questions
- How much VRAM does Whisper large-v3 need?
- For inference, Whisper large-v3 needs about 3.4 GB in BF16, 1.7 GB in FP8, 0.9 GB in INT4, including a 10 % runtime margin. Real usage depends on the inference engine and its settings.
- Can Whisper large-v3 run on a single GPU?
- Yes. In BF16 it needs about 3.4 GB, which fits one NVIDIA GeForce RTX 5080 16GB (16 GB).
- How much GPU memory does fine-tuning Whisper large-v3 take?
- QLoRA (4-bit base): about 1 GB; LoRA (BF16 base): about 4 GB; Full fine-tuning (BF16, AdamW): about 27.3 GB. Longer training sequences and larger batches add activation memory on top.
- What is the best GPU for Whisper large-v3?
- Best for inference: 1 GPU × NVIDIA L4 24GB (FP8 or INT8 weights, 1.7 GB). Best for fine-tuning: 1 GPU × NVIDIA L4 24GB (LoRA on BF16 weights, 4 GB). Budget pick: 1 GPU × NVIDIA GeForce RTX 5080 16GB (Smallest card class that holds a 4-bit build, 0.9 GB). Check live offers before booking, as availability changes.
- Can Whisper large-v3 run on an RTX 4090?
- Yes. In FP16 it needs about 3.4 GB, which fits one RTX 4090 (24 GB).
- Can Whisper large-v3 run on an RTX 5090?
- Yes. In FP16 it needs about 3.4 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.