Celeris-1: A diffusion LLM benchmarked at 2,082 output tokens/s

Tom here, CEO at Celeris. I'll be around all morning. Ask me anything about the model, diffusion LMs, or the numbers.