DEVOURED
Introducing celeris-1
The new celeris-1 model claims to match GPT-5 intelligence with 15x faster response times by utilizing a novel diffusion-based inference architecture.
What: Celeris-1, a general-purpose language model, reports a p50 response latency of 157ms and a throughput of 1,280 tokens per second, which the developers attribute to replacing standard transformer inference with a diffusion-based approach.
Why it matters: If confirmed, this suggests diffusion techniques—typically reserved for image generation—could become a viable alternative to autoregressive models for latency-sensitive text tasks.
Original article
celeris-1 is a general-purpose language model that delivers near-GPT-5 level intelligence with 15x faster response times. It uses a new inference architecture that uses diffusion techniques, unlocking dramatically better speed while maintaining frontier-level intelligence. The model delivers p50 response latency of 157ms and a throughput of 1,280 tokens per second. A link to a post on how the model was built along with full benchmarks is available.
Explore all 65 articles in today's digest
Fresh dev news, AI-digested daily. Never miss a trend at devoured.fyi