All tasks / Models / Granite Speech 4.1 2B
Alternatives to Granite Speech 4.1 2B
For each task Granite Speech 4.1 2B is listed in, the highest-ranked other options of the same kind, ranked only from public benchmarks.
Transcription: models
A
Granite Speech 4.1 2B · Local · 6% on Hugging Face Open ASR Leaderboard (English, short-form average WER)
| Tier | Name | Approach | Runs | Rated by | Price | Skill |
|---|---|---|---|---|---|---|
| S |  Fun-Realtime-ASR (preview)Alibaba | Models | Cloud | 1.7% on Artificial Analysis Speech to Text: AA-WER v2 (non-streaming) | — | — |
| S |  MAI-Transcribe-2 (preview)Microsoft AI | Models | Cloud | 2% on Artificial Analysis Speech to Text: AA-WER v2 (non-streaming) | $0.0017/min of audio | — |
| S |  Scribe v2ElevenLabs | Models | Cloud | 2.2% on Artificial Analysis Speech to Text: AA-WER v2 (non-streaming) | $0.0037/min of audio | — |
| S |  Grok Voice Transcribe 2.0xAI | Models | Cloud | 2.3% on Artificial Analysis Speech to Text: AA-WER v2 (non-streaming) | $0.0017/min of audio | — |
| S |  StepAudio 3 ASRStepFun | Models | Cloud | 2.5% on Artificial Analysis Speech to Text: AA-WER v2 (non-streaming) | $0.0067/min of audio | — |
| S |  Gemini 3.5 TranscribeGoogle | Models | Cloud | 2.6% on Artificial Analysis Speech to Text: AA-WER v2 (non-streaming) | $0.0050/min of audio | — |
| A |  Ink-2Cartesia | Models | Cloud | 3.4% on Artificial Analysis Speech to Text: AA-WER Streaming | $0.0040/min of audio | — |
| A |  Qwen3-ASR-1.7BAlibaba (Qwen) | Models | Local | 5.4% on Hugging Face Open ASR Leaderboard (English, short-form average WER) | ~6 GB VRAM | — |
Tiers come only from public benchmarks; tools without a result of their own inherit the tier of the best model they offer. How we rank.