Grok

33.1 As usual experience score ▲ 0.9 Confidence: High

319 opinions tracked this week · 7-day rolling window

30-day experience score

102030405060702026-07-31 · Score dropped 21 pts vs prior 7-day average2026-08-01 · Score dropped 16 pts vs prior 7-day average2026-08-02 · Score dropped 14 pts vs prior 7-day average2026-08-03 · Score dropped 12 pts vs prior 7-day average2026-08-04 · Score dropped 13 pts vs prior 7-day average07-2808-0408-1208-1908-26

Version breakdown

Only comments that explicitly name a version contribute to its score.

Grok 4.6 74.4+41.3 n=28 Confidence: Medium
Grok 4.5View lifecycle → mentions only n=4 Confidence: not enough data

289 generic family mentions are included in the family score, not any version score.

By experience dimension

General text 36.4 n=106
Coding 74.8 n=20
Reasoning 37.9 n=26
Image / vision 16.9 n=56
Video generation 30.9 n=5
Roleplay / creative 57.8 n=27
Safety & refusals 16.8 n=61
Speed & latency 56.8 n=19
Local deploy not enough data n=0

Top positive drivers

  • roleplay limit Roleplay / creative ×10
  • speed quality Speed & latency ×9
  • problem solved General text ×9
  • strong reasoning Reasoning ×8
  • competitor flight General text ×8

Top negative drivers

  • safety overreach Safety & refusals ×43
  • performance issue General text ×16
  • hallucination Reasoning ×12
  • creativity down Image / vision ×8
  • creativity down General text ×8

What users are saying

Selected from the latest 7 days, prioritizing recency, evidence quality, confidence, and engagement.

🔴 “I'm getting ghosting in 720 now too. It's ridiculous.”

via Reddit · Video generation · 2026-08-26

🔴 “Ghosting issues at 720/1080, recommends 480p”

via Reddit · Video generation · 2026-08-26

🟢 “I use a lot Grok 4.6 for UI adjustment and simple stuff that doesn't need a plan. It's incredible value even on a $20 plan. It's pretty token efficient even adding screenshots etc for context”

via Reddit · General text · 2026-08-26

🔴 “转deepseek和grok的时候同样也有打拳的问题”

via Zhihu · Safety & refusals · Chinese · 2026-08-26

🔴 “That's entirely Grok's fault though. They trained their models on content they don't want to be producing. If they don't want something done with it, don't include it in your training data.”

via Reddit · Safety & refusals · 2026-08-26

How's your AI experience today?

One tap, no sign-up. Pick a model and tell us.