Gladia's split speech workflows

This split shows Gladia is selling a menu of speech workflows, not one model that gets better in every direction at once. Solaria-3 is the cheaper, narrower engine for five-language business recordings, while Solaria-1 and Gladia’s live API still cover the broader jobs, like multilingual conversations, language switching mid call, and streaming captions or agent responses. That means customers still have to route each workload to the right endpoint instead of standardizing everything on the newest model.1 2 3 4

• In practice, this changes product design. A meeting app can use Solaria-3 for uploaded English sales calls, but if the same customer wants live captions, or a call that moves between French and Arabic, the app has to fall back to Gladia’s broader live or multilingual path.1 2
• The limitation is competitive, not cosmetic. AssemblyAI documents streaming diarization in its real time API, while Deepgram markets Flux around end of turn detection for voice agents. Gladia’s lack of live speaker labeling and Solaria-3’s recorded only scope leave gaps in the exact workflows buyers test first for call centers and agents.1 3 4 5
• It also affects monetization. Gladia lists lower starting prices for recorded Solaria-3 than for Solaria-1 or live transcription, which suggests Solaria-3 is built to win batch business recording volume, while broader multilingual and live jobs still carry premium pricing because they are harder to serve.1 2

The next step is likely a tighter product ladder, where Gladia keeps the low cost recorded lane for bulk business audio and expands newer models toward live, speaker aware, and more multilingual workloads. Winning more voice agent and enterprise call volume will depend on closing those workflow gaps without losing the price advantage that makes Solaria-3 attractive.1 3 4