Which AssemblyAI-STT-3.7 model should you use?
Blog post from AssemblyAI
AssemblyAI-STT-3.7 offers three models selected by whether audio is recorded or live and, for live audio, whether software responds: the pre-recorded model handles stored audio at $0.21 per hour, Realtime supports live transcripts and speaker labels for people to read at $0.30 per hour of session time, and Realtime-Agent supports voice agents with features such as Voice Focus and manual turn control at $0.45 per hour. All three cover 32 languages and share an API key; teams can combine them, such as using Realtime during a meeting and the pre-recorded model for its stored transcript. The Sync and Dictation APIs remain on Universal-3.6 Pro, with Dictation adding a rewrite step. The guidance also notes migration changes: some accounts switch to the new pre-recorded default on December 8, 2026, while several older streaming models have tentative retirement dates of February 1, 2027. Before migrating, users should account for differences such as disfluencies being retained by default and the need to list Universal-2 explicitly for fallback beyond the new model’s 32 languages.
| Trend | Post Mentions | Total Month Mentions | Posts | Companies | MoM |
|---|---|---|---|---|---|
| Real-time | 77 | 2,990 | 636 | 190 | -28% |
| Voice AI | 7 | 1,587 | 186 | 43 | -43% |
| LLM | 2 | 3,355 | 742 | 191 | -49% |
| Universal-3 Pro | 2 | 5 | 4 | 1 | +150% |
| Universal-3.6 Pro Realtime | 2 | 115 | 20 | 2 | -52% |
Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.