@lord.luke5: Dog because of this, I have created a new community so that you guys can basically you know learn about all these Chinese models that are now at the fable level, which is insane #china #Tech #Kimi #glm #Tech

Luke Luck engineer
Luke Luck engineer
Open In TikTok:
Region: US
Sunday 19 July 2026 20:48:49 GMT
6772
218
41
13

Music

Download

Comments

megasxlr31
Ryan :
Has anyone have a real codeing/ thinking test besides a copy paste benchmark? How does it do on code it’s never seen?
2026-07-20 21:02:36
0
cars30570
Car :
A lot of the times these Chinese models are just good at benchmarks and fall short in actual agentic use
2026-07-20 01:43:39
14
yilongmaaa
Yi Long Ma :
Claude is the best, that’s why everything is being compared to it. I’m happy to see cheaper alternatives though
2026-07-20 01:53:48
8
larry123210
Larry :
Claude guys gotta pay 200$/month so they gotta justify it somehow by conivincing themself it’s sooo much better. Cuz if they don’t convince themselves then they’ll feel bad about the price and feel they got duped for being impatient . lol
2026-07-20 10:51:26
0
dimchig_
DimChig :
nobody gonna have 10 gpus at their home bro
2026-07-20 03:01:40
1
drydogg11
gmoney :
Why would you say “almost fable level” for some and “better than opus” for others. The lack of quantitative metrics is a bad look
2026-07-19 22:05:33
3
naviwinn_
Naviwinn_ :
K3 is not almost fable level
2026-07-20 05:06:20
1
deja_marx
Chris :
Looking forward to Qwen3.8. 3.6, in both the 27B and 35B with MTP has been so good
2026-07-20 05:50:02
2
wwf7105
wwf :
Lies
2026-07-20 02:07:07
0
realfreetalk1
B J :
as a claude user its a great day, I have more options now
2026-07-20 00:35:18
4
romereezmjk
Romero :
2026-07-19 23:55:46
1
citizencapet
capet :
Falling behind lol. Pretty much cope. I use fable 5 ultra mode. I’m not falling behind. I like
2026-07-19 23:58:52
0
shortone2026
shortone2026 :
Deleted my last comment since I said the wrong mode name. You a fan of Sam Ovens?
2026-07-20 00:04:53
1
delgatto_x
delgatto_x :
we should be at a higher abstraction layer, the model shouldn't matter
2026-07-20 01:57:25
1
therealargonate
therealargonate :
As much as I wish this was true your comparisons are way off. Maybe at best better than Opus 4.6
2026-07-20 13:17:29
0
kaio___ken
KaioKen :
So you get to rent your GPU. And use the Chinese model!
2026-07-20 02:23:59
0
jlund2022
Jim Lundy :
Have a take….dont…..
2026-07-19 23:23:44
0
eabusham2
Eabusham2 :
Right.. like I can run it locally😂 I’d rather stick with codex for pricing and Claude for 5hour window
2026-07-19 22:18:33
0
wyhycu
The Tech Versatilist :
It’s easy for them to claim this when they run their own benchmarks. But when independent benchmarks are run it turns a bit different..
2026-07-20 00:47:17
1
hunterlamirande
Hunter :
GLM is actually not that good
2026-07-20 02:26:25
0
tonytimett
Tony Villalobos :
Why would it be a bad day 😂😂😂 I have Claude control them all and its still the best 😂👌 Makes my 20 dollar subscription worth 200 😁
2026-07-20 02:34:37
0
To see more videos from user @lord.luke5, please go to the Tikwm homepage.

Other Videos


About