@sustainment: Art by @Kedr.bit GPT-5.6 Luna was supposed to be a "cost-efficient" model, but when compared to DeepSeek V4 Flash, that claim starts to look questionable. V4 Flash is cheaper, freer, smaller, and open, while also matching or surpassing Luna in performance. What makes the comparison even more interesting is how similar some of Luna's capabilities are to V4 Flash, making Luna look less impressive in hindsight and more like an imitation of what DeepSeek accomplished. #fyp #deepseek #chatgpt #openAI #AI
_
Region: FI
Sunday 16 August 2026 12:13:44 GMT
Music
Download
Comments
ulei_de_granola :
luna on max beats all deepseek
2026-08-22 10:18:01
0
A Mysterious Figure :
Gemini is best anyways (they steal data tho)
2026-08-22 12:22:32
0
idktbh :
deepseek cost went up 4x today btw
2026-08-17 09:52:47
98
NotAnderx :
Just selfhost qwen 3.8 27b it’s genuinely close to as good as closed models like deepseek v4 flash
2026-08-16 19:00:32
50
Bameninghong :
until Gemini 3.7 Flash enters
2026-08-19 10:14:01
4
カネキ :
text-only 💪
2026-08-17 04:22:03
12
Ol :
after price increasing luna is cheaper per task than v4f, looks like they are competing straight with DS, making their chat app with Luna completely free and reducing its price by 80% on api
2026-08-16 18:10:03
26
domter :
this video aged like milk
2026-08-17 12:53:17
3
maxwell :
tt feed merging into my x feed slowly
2026-08-16 18:27:18
21
1rayhan._ :
you also get like unlimited usage on 5.6 Luna max with ChatGPT plus
2026-08-17 22:49:20
3
𐔌 . ⋮ star.aep .ᐟ ֹ ₊ ꒱ :
Kimi on top
2026-08-16 23:01:51
2
implication :
price updated btw
2026-08-16 19:03:30
22
idom :
deepseek price 2-5 timed itself yesterday and now Luna is in fact likely is a better choice
2026-08-17 10:38:30
1
sakupen09 :
idk about deepseek, been with anthropic for About a year now, anyone has any experience with both that could compare?
2026-08-16 13:24:17
4
︎ :
good that i dont pay API rates. subscriptions are far more generous. with a 20 usd sub you can actually get like 700 usd in compute if you are able to fully hit usage limits, but thats impossible with luna.
2026-08-16 20:03:30
1
Yukari :
Still no agent😭✌️🥀
2026-08-16 14:47:31
0
ulcerinmyleftnutsack :
I've used kimi and deepseek and like most other mainstream models, but God claude is so capable and smart, i mean in technical stuff it's so good at those, but sadly it's expensive and too many restrictions
2026-08-17 10:29:46
2
IniAdrian :
no flash v4 deepseek more expensive than luna
2026-08-20 01:35:41
0
dm1tryshyn :
recently made top-up on deepseek v4flash , after 2 hours of coding with swarm mode and high effort he spend like 30 cents , i was fucking shocked
2026-08-17 13:04:42
4
່ :
kedr.bit mentioned
2026-08-17 11:55:51
2
🧸 :
Deeply seeking non mid results on the actually relevant benchmarks
2026-08-17 02:26:05
4
SynastrIA Networks 🇧🇷 :
kedr mentioned 👀
2026-08-17 12:34:48
1
To see more videos from user @sustainment, please go to the Tikwm
homepage.