IBM Releases Granite 4.0 1B Speech, a Compact Model for Edge Devices

IBM has released Granite 4.0 1B Speech, a compact speech-language model designed for enterprise applications on edge devices. The model, released March 6, 2026, supports multilingual automatic speech recognition (ASR) and bidirectional speech translation (AST) in English, French, German, Spanish, Portuguese, and Japanese.

Despite having only half the parameters of its predecessor (granite-speech-3.3-2b), the new model delivers higher English transcription accuracy and faster inference through speculative decoding, according to IBM. It recently ranked #1 on the HuggingFace OpenASR leaderboard, which measures performance of open-source speech recognition systems.

Enterprise focus

The model is specifically built for resource-constrained devices—a key differentiator from frontier models that require significant compute. IBM recommends pairing it with Granite Guardian for production deployments that need additional risk detection.

New features in this release include Japanese ASR support and keyword list biasing, which improves recognition of names and acronyms—capabilities frequently requested by the community, according to the announcement.

Availability

Granite 4.0 1B Speech is released under an Apache 2.0 license with native support in Hugging Face Transformers and vLLM. The model achieves competitive Word Error Rate (WER) results on standard English ASR benchmarks while using far fewer parameters than comparable models, according to IBM's model card.

This release marks IBM's continued push into enterprise-focused AI, positioning Granite as an alternative to larger models for organizations with constrained compute budgets or strict data privacy requirements.