The Celeris blog

Notes on speed.

How we measure it, how we build it, and what a low-latency model unlocks. Written by the team building the fastest LLM on Earth.

July 8, 2026
Benchmarks · Engineering

How we test Celeris: speed and quality on the same footing

We ran Celeris against GPT-5, GPT-5 mini, and Amazon Nova Micro on MMLU-Pro and a real-world JSON workflow suite — one identical harness, quality and latency from the same run. The methodology, the numbers, and how we separated model time from the network.

More soon.