← Back to Voice models
Deepslate
Deepslate Opal
Release date unpublished
Deepslate's proprietary end-to-end speech-to-speech model, built in Germany and hosted in the EU or on-premises for GDPR-sensitive voice agents.
- Price per minute
- $0.10/min
- Inputs
- audio
- Tool calling
- Yes
- Languages
- 27
- Voices
- —
- Big Bench Audio
- 85%
- Access
- Generally available
- Licence
- Proprietary
Strengths & weaknesses
Strengths
- 0.44s time to first audio, which Deepslate cites from AA
- 27 languages with EU or on-premises hosting
- Function tools and webhooks for live lookups
Weaknesses
- Big Bench Audio 85.0% (AA)
- No published release date; v3.0 is a preview that requires access approval
Not notable: #23 on AA Big Bench Audio (85.0%). No release date on Deepslate's site or docs. The docs list opal-v2.1 (GA, platform default) and opal-v3.0-preview (on request). Price is the regular rate; a release offer of 2 ct/min was running. Deepslate's own page says the reasoning core is based on an open-weights LLM.
Links
API model strings
- deepslate
opal-v2.1