Developing next-generation accelerators requires benchmarking them on realistic and agentic AI workloads.
As part of the ARIA Scaling Compute programme, we are releasing TEAS, serving 5 models (4B to ~1T total) on 9 accelerators for 6 workloads. @epcc.bsky.social
teasbench.com
teasbench.com
TEAS Benchmark
TEAS reports measured cost, accuracy, performance and energy for LLM inference across models, workloads, engines and accelerators.
