Nerf Watch

ClaudeCommunityby tokenusage.dev

Nerf Watch gives the same private, code-graded tests to nine Claude and GPT models every day: Claude models through Claude Code on a Claude Max plan, and GPT models through Codex on ChatGPT Pro. Each model is compared with its own recent baseline. This connector reads the published results through two read-only tools. get_nerfwatch_scores returns each model's latest score, previous runs, change and baseline. get_nerfwatch_history returns one model's retained daily score history, oldest first. Use it to answer questions such as "Has Opus 5.5 scored lower this week?" or "How has GPT-6.1 Sol's score moved since October 20?" Limits: each model runs once a day, so a single day is noisy. Missing days are missing, not zero, and comparisons hold within one suite version. A lower score is a lower score on this suite, not proof that a provider changed a model. The tests themselves stay private. No account or sign-in is needed, and the connector writes nothing. tokenusage.dev is independent and not affiliated with Anthropic or OpenAI.