Tell-1 is Retell's own LLM, built for contact centers. It ranks #1 on cost, intelligence and latency for contact center calls, and it reasons while it speaks.
- Interleaved reasoning: most voice agents pause to think. Tell-1 answers right away and works out the next step while it talks: which rule applies, what is still missing, and when to call a tool.
- Top average on the Retell vCX benchmark: 94.4% across instruction following, grounding, tool calling and colloquial speech. It places in the top five on all four skills and first for sounding natural on the phone.
- Fast and affordable: about 150 ms to first token, at $0.035 per minute.
- Built for real SOPs: a contact center can have a thousand intents, each with its own procedure. Tell-1 finds the right one, follows it step by step, and brings the caller back when they drift.
- Independent benchmarks: results from Cekura, Bluejay, Hamming and Coval, plus public benchmarks like τ²-bench, are on the Tell-1 page.
- Availability: Tell-1 is in beta and runs on Retell's own inference stack. Select it as the LLM for your agent in the dashboard.
→ Learn more about Tell-1
→ Read the technical blog
→ See the benchmarks



