Current ranking
General text
For the 7-day window ending 2026-09-13 21:17:03 UTC, user-experience leaders in General text: closed models — Claude Opus 4.6 (72.7/100, n=8); open models — GLM 5.2 (73.2/100, n=14).
#1 Claude Opus 4.6n=8 72.7▾
“Opus 4.6 was the last good model that could hold a conversation without devolving in to being an asshole while still being intelligent.”
via Reddit · 2026-09-12
Open model detail →“massive love-fest for Opus 4.6. Y'all think it's the GOAT, praising its direct, sharp, and "human" personality”
via Reddit · 2026-09-11
#2 Claude Fable 5.1n=35 62.4▾
“Fable 5.1 is still my go to”
via Reddit · 2026-09-13
Open model detail →“chewed through my usage for a few turns for sure-- honestly it was pretty nuts”
via Reddit · 2026-09-12
#3 GPT-5.6 Lunan=11 61.1▾
“I'm using a lot of gpt-5.6-Luna”
via Hacker News · 2026-09-12
Open model detail →“right now just gpt Luna is enough I swear”
via Reddit · 2026-09-12
#1 GLM 5.2n=14 73.2▾
“it's just that one—like GLM 5.2—charges me $3, whereas DeepSeek v4-flash-0731 did the same job for a ridiculously lower price.”
via Reddit · 2026-09-12
Open model detail →“glm5.2 连你们所谓的炼炸的 dspro 都不过”
via Zhihu · Chinese · 2026-09-10
#2 Gemma 4 31Bn=10 71.9▾
“gemma4:31b”
via Reddit · 2026-09-13
Open model detail →“I still use Gemma 4 31B for technical writing. With a tiny bit of prompting, it's a breath of fresh air compared to the junk that Claude, GPT-5.6-sol, and Qwen3.8 27B are outputting these days.”
via Reddit · 2026-09-09
#3 GLM 5.3 Flashn=66 68.2▾
“glm flash seems like it doesnt use the memory context effectively”
via Reddit · 2026-09-13
Open model detail →“around the same level as 5.3-flash in my day to day”
via Reddit · 2026-09-13