Reddit proposes measuring LLMs by cost per accepted task, not cost per token, after Grok 4.7 launch
Crescitaly · reddit · 2026-09-22
xAI announced Grok 4.7 on September 21, claiming better long-running work and self-checking while keeping Grok 4.6's base price and speed.
The author argues that for research and marketing work, leaderboards matter less than a different comparison: give each model the same source-backed brief, then count how much human cleanup is needed—unsupported claims, missed requirements, revision rounds, elapsed time, and total usage—with success criteria fixed in advance. The core point: a cheaper token rate doesn't tell you the cost of a completed task.
The author explicitly notes this is a proposed test, not an experimental result, and doesn't claim Grok wins it.
More from Models
- Robot build test: GPT-6 Astra stabilizes two-board structure, Claude Fable 5.2 falls short — qinzytech · 2026-11-12
- Zvi dissects Anthropic's latest system card, disputing the autonomy threshold call — Don't Worry About the Vase (Zvi) · 2026-11-12
- antirez backs low-bit DeepSeek 4.1 Flash over Qwen 27B dense on Apple M5 Macs — antirez · 2026-11-12
- Gemini re-renders images as text when vision is removed, dropping only 6% WSR — zacharynado · 2026-11-12
- Bare-harness coding benchmark: Fable 5.2 best overall, Gemini 4 Carbon best value — Semantic_meaning · 2026-11-12
- Google's Physical-Engineering's Last Exam: all 14 frontier models score below 50% — dl_weekly · 2026-11-12