Granite Speech 5.0 2B
- Price
- Free
- API
- No
- Weights
- Yes
WHAT IT DOES
Granite Speech 5.0 2B is IBM's speech model released October 21, 2026, under Apache 2.0. It transcribes English, French, German, Spanish, Portuguese and Japanese, and translates between English and those languages, plus English into Italian and Mandarin. IBM says a prompt can name the input language, ask for punctuation, and bias a list of up to 2,000 keywords. On the Open ASR Leaderboard's public English short-form sets, IBM reports an average word error rate of 5.06 percent and calls that second among top open-weight speech models as of October 21, 2026. It runs in Transformers, vLLM, llama.cpp and mlx-audio.
WHY WE PICKED IT
Open speech models that handle more than English, and that you can run yourself, are still rare. This one is Apache 2.0, two billion parameters, and IBM publishes the word-error figure on the public sets rather than only a private mix.
QUICK FACTS
STRENGTHS
- ✓Apache 2.0 weights at 2 billion parameters
- ✓Transcription in six languages and translation to and from English
- ✓Keyword lists of up to 2,000 terms
- ✓Average English word error rate of 5.06 percent on public short-form sets, per IBM
- ✓Runs in Transformers, vLLM, llama.cpp and mlx-audio
LIMITATIONS
- ⚠A speech model, not a general chatbot
- ⚠The 5.06 percent word error rate is IBM's average on public English short-form sets
- ⚠Translation into Italian and Mandarin is English-to-those-languages only
- ⚠No IBM-hosted API is part of this release. You run the weights
GETTING STARTED
Download ibm-granite/granite-speech-5.0-2b. IBM's usage examples are on the model card. Name the spoken language in the prompt when you know it, and pass a keyword list for names and product terms. For speed on device, IBM points to the smaller granite-speech-5.0-470m-turboctc model instead.
OFFICIAL LINKS
SIMILAR TOOLS
PRICING
FEATURED IN GUIDES
What is Text-to-Audio AI? Complete Guide 2026
Text-to-audio AI generates voice, music, and sound effects from text descriptions. How AI audio tool...
What is AI Voice Generation? Complete Guide 2026
AI voice generation creates natural-sounding speech from text. How AI voice synthesis tools generate...
What is AI Music Generation? Complete Guide 2026
AI music generation creates original music from text descriptions. How AI music tools compose melodi...
How to Use Text-to-Audio AI Tools: Complete Guide 2026
Text-to-audio AI tools for music, voice, and sound generation. Prompt engineering, model capabilitie...
AI Music Production Tools: What Producers Actually Use
Discover the best AI tools for music production: Suno, Stable Audio 2.5, and other AI music generato...
EXPLORE ALTERNATIVES
Compare Granite Speech 5.0 2B with 5+ similar text → audio AI tools.
FREQUENTLY ASKED QUESTIONS
Is Granite Speech 5.0 2B free?
Yes, Granite Speech 5.0 2B is completely free to use.
Does Granite Speech 5.0 2B have an API?
No, Granite Speech 5.0 2B does not appear to offer a public API.
What is Granite Speech 5.0 2B best for?
Granite Speech 5.0 2B is best for Best Open Speech Model. Granite Speech 5. Open speech models that handle more than English, and that you can run yourself, are still rare. This one is Apache 2.0, two billion parameters, and IBM publishes the word-error figure on the public sets rather than only a private mix.
What platforms does Granite Speech 5.0 2B support?
Granite Speech 5.0 2B supports local.
Is Granite Speech 5.0 2B open source?
Yes, Granite Speech 5.0 2B is open source. You can access the source code, contribute, and deploy it on your own infrastructure.
How do I get started with Granite Speech 5.0 2B?
Download ibm-granite/granite-speech-5.0-2b. IBM's usage examples are on the model card. Name the spoken language in the prompt when you know it, and pass a keyword list for names and product terms. For speed on device, IBM points to the smaller granite-speech-5.0-470m-turboctc model instead.
How do I generate audio with Granite Speech 5.0 2B?
Granite Speech 5.0 2B creates audio from text descriptions. Enter prompts describing the type of audio you want (voice, music, sound effects) along with style, tone, and duration details.
FEATURED ON CURATEDAI
Work on Granite Speech 5.0 2B? You're hand-reviewed in our directory. Add this badge to your site — it links back to this profile.
