Whisper Base

Compact Whisper checkpoint for lightweight multilingual transcription and translation where modest hardware matters more than peak accuracy.

OpenAI · Whisper

Editorial review

Reviewed byOpenSourcesAI EditorialLast updatedAugust 2026SourcesExact model card

Model checkpoints, context windows, provider support, local runtime compatibility, and license terms can change quickly. Verify the exact model card before production or commercial use.

Best for

Developers who want a practical starting point for local captioning, voice notes, and CPU-leaning ASR experiments.

Who should use it

Developers who want a practical starting point for local captioning, voice notes, and CPU-leaning ASR experiments.
Builders who want local or self-hosted testing options.

Common workflows

CPU-friendly transcription experiments
audio workflows
transcription workflows
speech recognition workflows
local workflows

Deployment and hardware notes

~0.15 GB in fp16 — it runs comfortably on a laptop CPU and inside memory-constrained containers.

License and usage notes

Apache 2.0. Open source weights and code. Verify the exact model card and license terms for the checkpoint or hosted provider you use.

Strengths

Open source weights and code model option for Whisper workflows.
Developers who want a practical starting point for local captioning, voice notes, and CPU-leaning ASR experiments.
A sensible first checkpoint when wiring up an audio pipeline: get the plumbing working here, then swap in a larger size without changing any code.

Limitations

Accuracy is prototype-grade: expect misheard proper nouns and dropped words on anything but clear, close-mic English. Not a production transcription choice.
~0.15 GB in fp16 — it runs comfortably on a laptop CPU and inside memory-constrained containers.
Context window and limits: 30-second audio windows · 448-token cap per window.
Verify the exact model card, provider docs, license, and serving support before production use.

Local workflow notes

A sensible first checkpoint when wiring up an audio pipeline: get the plumbing working here, then swap in a larger size without changing any code.

Local runtimes: Transformers, faster-whisper, whisper.cpp

Platforms: Windows, macOS, Linux

Transcription spec

Memory~0.15 GB in fp16 (73M parameters) · ~0.07 GB in int8Decoder depth6 layersLanguagesMultilingual (99 languages)Audio window30-second audio windows · 448-token cap per window

6 encoder and 6 decoder layers. Fast enough for real-time transcription on modest CPUs, and small enough to ship inside an application rather than served alongside it.

Sources to verify

Exact model card

Related resources

Continue with model source notes, local tools, and implementation guides related to this model.

Whisper family hub AI tools directory Model source notes Models directory Guides library

Hardware~0.15 GB in fp16 (73M parameters)RuntimeTransformers, faster-whisper, whisper.cppContext30-second audio windows · 448-token cap per windowLast updated2026

Exact model card →