Deepgram review

Input price from $0.30 and output price from $0.60 are Deepgram's entry rates. Both sit above the €0.08 and €0.23 medians for this group, so short runs cost little while steady volume costs more than gateways built around text models.

New signups receive a free credit and no credit card is required to claim it. The entry rates behind that credit are $0.30 for input and $0.60 for output, so send a sample recording, check accuracy on your own audio, and only then decide whether ongoing volume is worth it.

The published set in Flux English, Flux Multilingual and Nova 3 Monolingual covers both directions of speech work. Flux and Nova handle English and multilingual recognition, the Whisper entries add an open-model path, and Flux TTS with the Aura models cover synthesis on the same account.

The largest context window listed is 150,000, which ranks 60 of 72 brands on that measure. The group median is 1,000,000, so teams feeding very long transcripts should check the window before committing.

The published list is SOC 2 Type 1 & Type 2, HIPAA and GDPR. Those cover the frameworks regulated buyers ask about first, and certification scope can differ by product, so confirm the exact product and your own region's requirements with the vendor before signing.

Support runs on help centre and community: a help centre and a community. Phone and live chat are not among the listed channels, so teams that need a named contact on a call should plan for documented answers instead.

Deepgram stays inside speech, with Flux English, Flux Multilingual and Nova 3 Monolingual as its published catalogue. Salad's input price from $0.06 is lower for a raw entry rate, Azure AI Foundry lists general-purpose models in gpt 6 astra, gpt 6 luna and gpt 6 sol, and Supascale AI gathers open families in DeepSeek V4 Pro, GPT 5.6 Sol and GPT 5.6 Terra.