Current ranking
Reasoning
For the 7-day window ending 2026-09-13 23:17:02 UTC, user-experience leaders in Reasoning: closed models — Claude Fable 5 (82.3/100, n=64); open models — Qwen3.8 27B (80.0/100, n=45).
#1 Claude Fable 5n=64 82.3▾
“Astra is very, very good. So is Fable. I think we're still getting a feel for what each is better at. There are definitely use cases where Astra beats Fable and vice versa”
via Reddit · 2026-09-13
Open model detail →“Fable ends up saying Astra is right and it missed x because it had either made some assumptions or not gone deep enough”
via Reddit · 2026-09-13
#2 GPT-6 Astran=189 81.0▾
Open model detail →“Astra is no match for very architecturally or algorithmically complex task”
via Reddit · 2026-09-13
#3 Claude Fable 5.1n=25 74.4▾
“I think Fable 5.1 was nerfed and I cannot trust it anymore on this kind of logic”
via Reddit · 2026-09-13
Open model detail →“Fable seems to forget about things”
via Reddit · 2026-09-10
#1 Qwen3.8 27Bn=45 80.0▾
“doesn't overthinking like 3.8 does on xhigh”
via Reddit · 2026-09-13
Open model detail →“Qwen3.8:27b doesnt know when to not think. It overthinks things that are simple and thinks hard for big tasks too (appropriately).”
via Reddit · 2026-09-13
#2 GLM 5.3 Flashn=34 79.8▾
“The flash version is very capable and fits into the luna category. Feels a little more capable than luna imo but more verbose.”
via Reddit · 2026-09-13
Open model detail →“I think the best here is glm 5.3 flash, it has a very low hallucination rate”
via Reddit · 2026-09-13
#3 DeepSeek V4 Flashn=18 75.7▾
“DSV4 it will just keep on going until it finds the issue, or if it cannot, then it'll stop eventually and prepare a handover for a fresh session”
via Reddit · 2026-09-13
Open model detail →“never have issues with agents going rogue like this”
via Reddit · 2026-09-13