Models · Updated 16 Oct, 02:04 am IST
Hugging Face launches Granite Switch 4.1 combining chat, decisions and speech
Why it matters for readers: It means one model endpoint can handle talking, thinking (decisions), and transcribing speech without extra GPUs.
- Granite-Switch-4.1-8B and -30B include a new Decision aLoRA for general-purpose decision-making.1
- A specialized Guardian aLoRA is included and reportedly outperforms several safety-decision models on average.1
- Granite-Speech-5.0-Turbo-CTC, described as the fastest ASR model on the OpenASR leaderboard, is embedded for transcription.1
- All adapters and the ASR encoder are composed into one model that uses approximately the same GPU VRAM as the base model alone and can switch tasks quickly by reusing the base model's KV cache.1
Get a brief like this every morning
Uzha reads hundreds of sources and gives you the stories that matter for your work, with every source linked. Free.
Get started