Granite Speech 5.0: IBM's open Apache-2.0 speech models for transcription, translation, and speaker tracking
In today's Spotlight · Picked by 1 of 4 modelsTurns speech into text, translates it, and tracks who is speaking.Like a multilingual note-taker for conversations.
IBM released granite-speech-5.0-2b and updated granite-speech-5.0-470m-turboctc and granite-speech-4.1-2b-plus — covering high-accuracy multilingual transcription and translation, ultra-fast on-device transcription, and multi-speaker attribution. All three are Apache 2.0 with native Transformers support.
Why it matters
One Apache-2.0 speech family spanning laptop to datacenter: #2 open-weight model on the Open ASR Leaderboard at 5.68% WER, runs on Transformers, vLLM, llama.cpp and mlx-audio with no per-token API cost.
What you could build with it
An indie developer building a privacy-first meeting-notes app could run granite-speech-5.0-2b locally on a laptop to transcribe and translate calls offline, using the 470m-turboctc model for near-instant live captions — a product with no per-minute API bills and no audio ever leaving the device.
Does it hold up?
Strong on paper, thin on independent evidence: the family covers distinct real deployment constraints (laptop, edge, meetings) and the Turbo CTC model reports 19,870 RTFx, but the accuracy lead (#2 on the Open ASR Leaderboard at 5.68% WER) is IBM-reported and no independent re-tests have surfaced yet.
Built with Granite Speech 5.0
- IBM updates Granite Speech with models for translation, speed and speaker trackingRuntimeWire analysis · RuntimeWire covers how IBM packaged speech AI around deployment constraints — separate models for multilingual translation, high-throughput transcription, and speaker tracking — and notes the reported speed/accuracy figures still need real-world testing.
- Granite Speech 5.0: local transcription and translationiMasters coverage · iMasters highlights the on-premise angle: granite-speech-5.0-2b runs on vLLM, llama.cpp and mlx-audio for Apple Silicon, so transcription and translation can run entirely on a laptop without sending audio to a third-party API or paying per processed voice token.
- Granite Switch 4.1: One 8B Model, Chat + Decisions + ASRExplainX analysis · ExplainX's breakdown of IBM's Granite Switch 4.1 notes it composes the Granite-Speech-5.0-Turbo-CTC ASR into the same endpoint as chat, decision, and safety adapters — one real example of the speech models being wired into a multi-skill deployment.
Learn more
First spotted on hf: source.
More AI models
OpenAI publishes 722 AI-generated math manuscripts from an unreleased modelOpenAI released 722 mathematical manuscripts in 372 result families to GitHub under Apache-2.0, produced by an…models · JEV 0.74Tencent Hunyuan Hy3: open 295B MoE with 21B active under Apache 2.0Tencent open-sourced Hy3, a 295B-parameter mixture-of-experts model with only 21B parameters active per token, under…models · JEV 0.73NVIDIA Isaac Lab 3.0 GA + Isaac OS open runtime + Agile ONE SN humanoidMajor Isaac platform expansion at GTC Berlin Oct 21: Isaac Lab 3.0 is generally available as the open-source robot…models · JEV 0.73StepFun Step-5-Preview — open BF16 weights releasedStepFun published the full BF16 weights for its flagship Step-5-Preview agentic model (600B-parameter sparse MoE, 27B…models · JEV 0.73NVIDIA Nemotron 3.5 Retriever collection: Embed 8B/1B, Rerank 8B, Parse 631MNVIDIA's Oct 21 collection drop for agentic retrieval: Nemotron-3.5-Embed in 8B BF16, 1B BF16 and 1B NVFP4…models · JEV 0.72Kimi K3.1 Pro: Moonshot's 2.8T open-weight flagship officially launchesMoonshot AI officially launched Kimi K3.1 Pro on Oct 23, a mixture-of-experts model with 2.8 trillion total / 104…models · JEV 0.72
Get the week's best AI launches, plus 3 ideas worth building
One email every Saturday. Ranked by traction, not hype. Free.