Inference for realtime agents.

Hopper trains models on your production traces, then serves and improves them against your live traffic

Talk to an Engineer
Zero data retention. By default.
Hopper Inference circuit board

Gemma 4 31B

80 ms

TTFT

vs. 600 ms for GPT-4.1

$0.30

per 1M input tokens

vs. $2 for GPT-4.1

Meet the team

We’re Pavan and Jashwanth. Pavan ran voice inference at Salient, powering millions of calls a day. Jashwanth wrote inference kernels for Quest 3 at Meta Reality Labs.

We met in high school more than 13 years ago and studied together at IIT Madras.

  • Meta
  • ACM ICPC
  • Uber
  • Microsoft
  • Samsung
  • IIT Madras

We’re backed by Y Combinator and leaders including Vijay Krishnan (CTO of Turing), as well as others from xAI, Stripe, and Rubrik.

Writing