// Whisper · Speech-to-text
whisper-large-v3: VRAM & local deployment
1.54B parameters. About 1.5 GB at 4-bit and 3.9 GB at BF16.
released 2023-11-07licence: Commercial OKmemory figures are estimates
01Specs
| Publisher | OpenAI · openai |
|---|---|
| Family / series | Whisper · whisper-large |
| Type | Speech-to-text (automatic-speech-recognition) |
| Parameters | 1.54B Source: counted from official safetensors weights |
| Released weights | FP16 · 9.3 GBdownload |
| Licence | Apache-2.0 · Commercial OKPermissive licence (Apache-2.0 / MIT / CC-BY / OpenMDW): commercial use allowed. |
| Released | 2023-11-07 (Hugging Face repo created) |
02How much VRAM it needs (estimate)
Speech models are small; the weight size is close to the total memory needed.
| Precision | Weights | Weights + overhead | Smallest GPU | Mac |
|---|---|---|---|---|
| BF16 / FP16 (original) | 3.1 GB | 3.9 GB | 8GB GPU | Mac 16GB |
| FP8 / INT8 / Q8_0 | 1.6 GB | 2.3 GB | 8GB GPU | Mac 16GB |
| 4-bit (Q4_K_M / AWQ / GPTQ-Int4) | 0.9 GB | 1.5 GB | 8GB GPU | Mac 16GB |
Bits per weight: BF16 16, FP8 8.5, 4-bit 4.85. “Fits” = total ≤ 90% of memory; Mac GPU can use ~2/3 of unified memory (≤36GB) or ~3/4 (larger). For MoE, all experts must be in memory, so size by total parameters. Method →
03How to run it
Only runtimes the publisher or the runtime’s official library names for this model.
- Transformers — mentioned in the official model card
04Official repositories & quantised builds
1 official repo(s) from openai.
- whisper-large-v3 main
05Other Whisper models
All 2 →| Model | Params | 4-bit total | Released |
|---|---|---|---|
| whisper-large-v3-turbo | 809M | 1.0 GB | 2024-10-01 |
06Sources
Data checked 2024-08-12. Parameters, context and licence are copied from the official repository; memory figures are our estimates.