Comment by esafak

If anyone from Meta is reading, please can you publish the cost and latency for each of your benchmarks, like OpenAI does? Show us how the reasoning effort level affects them in 2D charts. This needs to become standard practice.