Grade a delivered job. Code checks what it can. Jev answers the rest, and the answers are recorded.

gpu.inference.metered (inference.metered): POST /api/coordination/chat/completions {"max_tokens":16,"messages":[{"content":"What is 17 multiplied by 24? Answer with the number only.","role":"user"}],"model":"deepseek-v4-flash","prompt_tokens":14}

ok: true · id: chatcmpl-gpu_mviatjzt_99d88ec2a6 · object: chat.completion · 12 more fields