DeepSeek's super-agent is 1.9x faster — but burns 3.8x the tokens
AIFlow_ML · x · 2026-11-03
AIFlow dissects a DeepSeek super-agent test writeup: tasks finish 1.9x faster, but the article's own table shows 3.8x token consumption — the speed comes from burning vastly more tokens. The author also argues the coding task in the test was just a time sink, and breaks down what the numbers actually measured and which practices are worth copying.
Related event: DeepSeek Super-Agent Is 1.9x Faster but Burns 3.8x Tokens(4 posts)→
More from Models
- Dev compares astra and sol to Opus 5.5: rigid on open-ended tasks — haider1 · 2026-11-03
- Open-source models now beat every 'frontier' model from just 11 months ago — Hesamation · 2026-11-03
- Sebastian Raschka Maps October's 'Open October': 10 Open LLM Architectures in One Month — rasbt · 2026-11-03
- First hands-on: Meta's Muse Wonder 1.0 vs DeepSeek V4.1 Flash in Codex CLI — Richidm · 2026-11-03
- "Claude Code has been terrible today" — developer smells a new Opus launch coming — hermanschutte · 2026-11-03
- Mistral releases open-weight ML4 with frontier code-agent and cyber capabilities — _AndrewZhao · 2026-11-03