Whisper small: memory fit on every Jetson module.

Model facts

OpenAI · hf:openai/whisper-small · config.json · verified 2026-09-07

“"d_model": 768, "encoder_layers": 12, "decoder_layers": 12, "encoder_attention_heads": 12, "decoder_attention_heads": 12”

Parameters source

• OpenAI · github:openai/whisper/README.md · verified 2026-09-07 · openai/whisper GitHub README.md · class A

  “| small | 244 M | `small.en` | `small` | ~2 GB | ~4x |”

Published quantised artefacts

A file the publisher or a community mirror actually ships, used as the class-A weight figure when the requested quant matches. Any quant without a row here falls back to parameters × bytes-per-parameter (class D for block quants).

No published quantised artefact for this model; weights are estimated as parameters × bytes-per-parameter for every quantisation.

Jetson tokens/s measurements

Published or archived throughput numbers, not modelled. Class C (external measured benchmark).

No Jetson-measured tokens/s reported for this model yet.

Every module × every quantisation

Verdict and total memory at 4096-token context, 1 sequence, llama.cpp, headless. Q4 cells link to the static breakdown page; every other cell links to the live engine at that quantisation.

Notes

Encoder-decoder architecture: num_hidden_layers/hidden_size/heads above are per-stack (encoder_layers==decoder_layers==12, encoder_attention_heads==decoder_attention_heads==12 for this size). max_position_embeddings=448 is max_target_positions (decoder text token limit); encoder audio context is max_source_positions=1500 (30s of 20ms mel frames). torch_dtype in config.json is float32 (not bf16) -- params computed from OpenAI's published table, not from safetensors size, per instructions (parameters only for Whisper). Reference VRAM requirement '~2 GB' from OpenAI README is a generic estimate, not Jetson-measured.

Method and data: Model Memory Fit methodology. Full registry: model-memory.json.