Hopper
Voice model API: trains speech-to-text, text-to-speech and speech language models on a customer's production calls, then serves and improves them against live traffic; the home page shows a Gemma 4 31B language model at 80 ms time to first token and $0.30 per million input tokens, set against GPT-4.1; start in the console or talk to the founders.
What it does
- Model inference
- Model training
- Speech to text
- Text to speech
Features 3
- Trained on your callsTrains speech models on production calls and serves them on live traffic.
- Fast voice inferenceBuilt for latency-sensitive voice workloads.
- Token pricing shownShows $0.30 per million input tokens against GPT-4.1 at $2.
Sources
- Pricing
-
Official documentation · 18 Oct 2026
Home page: $0.30 per 1M input tokens, set against $2 for GPT-4.1; no plan table is published on the pages read.
- Description
-
Official documentation · 18 Oct 2026
Home page: "Hopper trains STT, TTS, and speech LLMs on your production calls, then serves and improves them against your live traffic"
- What it does
-
Official documentation · 18 Oct 2026
Home page: "Hopper trains STT, TTS, and speech LLMs on your production calls, then serves and improves them against your live traffic"
- Sold within
-
Console sign-in from Get started, with a Talk to an Engineer call offered; a per-token rate is published.
- Name
-
Official documentation · 18 Oct 2026
Name as published on the product page.
- URL
-
Official documentation · 18 Oct 2026
Product page, read 2026-10-18.
- Deployment
-
Official documentation · 18 Oct 2026
The pages read state no deployment model beyond the vendor-run service; saas is recorded as the vendor-operated reading.
- Interfaces
-
Official documentation · 18 Oct 2026
Terms of service: "our API, our console, our demos"; the home page links a console.
Something wrong here? Send a correction — quote this product id: hopper-voice-inference.