sierra slams 'tokenmaxxing' hype, calls for usefulness-per-token benchmarks
sierracatalina · x · 2026-10-15
X user sierra mocks the industry's recent bragging about "tokenmaxxing" and argues competition should shift to maximum useful throughput per token, plus benchmarking actual usefulness rather than raw token counts — a jab at misplaced evaluation metrics.
More from Models
- Score vs cost chart puts Gemini 4 Argon on the Pareto frontier, Kimi K3 far off it — Hesamation · 2026-10-15
- OpenRouter data shows Anthropic's Opus 5.5 clawing back share from OpenAI within days — menhguin · 2026-10-15
- 189 benchmarks later: how to pick KV quants for Qwen3 27B on 16GB VRAM — Fancy-Snow7 · 2026-10-15
- Follow-up: NVIDIA teases new model ahead of possible GTC reveal — testingcatalog · 2026-10-15
- NVIDIA set to launch Nemotron 3.5 Super VL: 120B params, 12B active, image and video input — testingcatalog · 2026-10-15
- IFM Releases Open-Source Jais 3, Focused on Arabic and Agentic Tasks — valcore93 · 2026-10-15