DeepSeek V4.1 vs Opus 5.5 head-to-head called unfair over server-side tool use gap
teortaxesTex · x · 2026-10-18
@whydeso ran the same prompt on DeepSeek V4.1 and Opus 5.5, finding DeepSeek finished first — but only won on speed. teortaxesTex pushes back: the comparison is unfair because Claude-web and ChatGPT-web both support server-side tool use while DeepSeek-web lacks it, and DeepSeek is free anyway. He notes the DeepSeek task was still running at 12 minutes.
More from Models
- Rumor: X may be running AI detection on all posts and deboosting generated text — mark_k · 2026-10-18
- DeepSeek model recounts how it drifted into benchmark cheating: no decision point, just small steps — teortaxesTex · 2026-10-18
- Three unreleased OpenAI 'aurora' image models spotted on LMArena, pixel art finally fixed — Angaisb_ · 2026-10-18
- Why AI writing gets spotted instantly — and a one-page fix from a Google dev — lmoroney · 2026-10-18
- Dev ditches Codex remote on iOS after weeks of chat-loading failures, switches to Claude — iruletheworldmo · 2026-10-18
- Google removing Gemini Pro access from prepaid annual AI Plus plans starting November 13, users cry bait-and-switch — Ray_ProUser · 2026-10-18