Mastodawn

Ivan Fioravanti ᯅ (@ivanfioravanti)

평가(Eval)에서 thinking tokens가 결과와 소요 시간에 큰 영향을 주며, 모델이 과도하게 생각하는 경우를 벌점화하기 위해 평가별 최대 토큰 제한이 필요할 수 있다는 제안이다.

https://x.com/ivanfioravanti/status/2045134936206963142

#evals #thinkingtokens #llm #benchmark #ai

Ivan Fioravanti ᯅ (@ivanfioravanti) on X

Thinking tokens have an incredible impact on evals in terms of duration and results. Finding right balance is difficult. I think we should have a fixed max token per eval in some cases to penalize models that are thinking too much. No?

X (formerly Twitter)

sayzard Apr 17

Boris Cherny (@bcherny)

Opus 4.7이 더 많은 thinking tokens를 사용하게 되면서, 모든 구독자에게 이를 보상하기 위해 rate limit이 상향되었습니다. 모델 사용 방식과 서비스 정책이 함께 조정된 업데이트입니다.

https://x.com/bcherny/status/2044839936235553167

#opus47 #ratelimits #llm #thinkingtokens #ai

Boris Cherny (@bcherny) on X

Opus 4.7 uses more thinking tokens, so we've increased rate limits for all subscribers to make up for it. Enjoy!

X (formerly Twitter)

AI Daily Post Nov 19, 2025

OpenAI just rolled out GPT‑5.1‑Codex‑Max, a new Codex CLI that finishes a 24‑hour coding sprint on its own. With “thinking tokens” and autonomous debugging, it pushes LLM‑driven development to a new level. Curious how it works and what it means for open‑source tooling? Dive into the details. #GPT5_1CodexMax #CodexCLI #ThinkingTokens #AutonomousDebugging

🔗 https://aidailypost.com/news/openai-launches-gpt51codexmax-completes-24hour-coding-task-internally