@myr3san: but l kinda hope they catch us #fyp #sunki #niki #sunoo #enhypen

yagmur
yagmur
Open In TikTok:
Region: TR
Sunday 11 October 2026 13:48:48 GMT
7458
1546
66
218

Music

Download

Comments

nnlxgh_n
dada :
silmeden kaydedelim bari süper olmuş qq
2026-10-11 16:36:13
1
user3409853279
z :
SUNKI
2026-10-11 13:51:49
3
rcorvet
18 :
Daha çok video at boşladın
2026-10-11 14:57:01
1
fmnk.1
،،ѕυηι⁞⊹ ࣪ ˖ :
i love sunki
2026-10-14 22:13:01
0
gwoozei
blue :
slm
2026-10-11 15:38:51
1
v3xsios
￴ ￴ ￴ ￴ ￴￴ ￴ ￴ ￴￴￴ ￴ ￴￴ ￴ ￴￴￴ :
güzel olmuş beğendim tmm
2026-10-11 14:17:00
1
aylisins
aylin :
2026-10-11 13:52:43
1
junniouts
ros :
TUTACAAAAKK
2026-10-11 15:35:58
1
snxria
gaymur :
bayilzzim buna ben💜💜
2026-10-11 15:41:24
1
llyrtela
johnny :
sunki mi
2026-10-11 14:16:07
1
rcorvet
18 :
Bana da öğret
2026-10-11 14:55:54
1
nnlxgh_n
dada :
sen hep edit yap
2026-10-11 16:36:17
1
rcorvet
18 :
Fanınım
2026-10-11 14:53:40
1
rcorvet
18 :
Harika olmuş tabbbi ki
2026-10-11 15:00:45
1
llyrtela
johnny :
bayildim znm
2026-10-11 14:16:10
1
junniouts
ros :
pekı cok guzel olması pekı
2026-10-11 15:36:11
1
rcorvet
18 :
BU ŞARKI
2026-10-11 14:55:23
1
junniouts
ros :
hep vıdeo at
2026-10-11 15:37:23
1
nnlxgh_n
dada :
ah sunki
2026-10-11 16:36:23
0
snxria
gaymur :
2026-10-11 15:41:15
0
v3xsios
￴ ￴ ￴ ￴ ￴￴ ￴ ￴ ￴￴￴ ￴ ￴￴ ￴ ￴￴￴ :
oha sunkı
2026-10-11 14:16:49
1
rcorvet
18 :
Slm seni beğendim sevgilim ol
2026-10-11 14:56:16
1
user3409853279
z :
cok muhtesem
2026-10-11 13:52:09
1
nnlxgh_n
dada :
lütfen
2026-10-11 16:36:20
0
llyrtela
johnny :
2026-10-11 14:16:12
0
To see more videos from user @myr3san, please go to the Tikwm homepage.

Other Videos

Mistral just released Large 4, and Europe finally has a trillion-parameter open-weight contender. ML4 (Mistral's nickname: le Chonk) is a 1T Mixture-of-Experts with 49B parameters active per token. It's natively multimodal, reasons and uses tools in one model, and was trained on 3,800 Grace Blackwell GPUs in Mistral's own European datacenters. The preview API is live today at $1.36 per million input tokens and $4.18 per million output, and the weights are promised by the end of the month. On Artificial Analysis's DeepSWE 1.1, as charted by Mistral, it scores 62: just ahead of GLM-5.3 at 61, but behind Kimi K3 at 68. In a blind human review of coding quality it placed second, behind only Claude Opus 5. The cyber angle is the real pitch. It scores 82% on reproducing and patching a real vulnerability, the top score on that test, partly because closed models like Claude Opus 5.5 and GPT-6 Astra refuse the task. Mistral says it still refuses malicious cyber prompts more often than any other open model. The fine print: you can't download it yet, the post doesn't name a license, and most charts are Mistral's own picks. Also, 49B active doesn't make it small to host. All 1T parameters still have to fit in memory. Mistral says it beats every open model built in the US or Europe. If the weights land on time, I want to see independent coding numbers before calling it. Source: Mistral's Introducing Mistral Large 4 post, Oct 6, 2026. Independent briefing by CodenamePoshan, not a Mistral post. #MistralAI #OpenSourceAI #LLM #Cybersecurity #AICoding #AIEngineering
Mistral just released Large 4, and Europe finally has a trillion-parameter open-weight contender. ML4 (Mistral's nickname: le Chonk) is a 1T Mixture-of-Experts with 49B parameters active per token. It's natively multimodal, reasons and uses tools in one model, and was trained on 3,800 Grace Blackwell GPUs in Mistral's own European datacenters. The preview API is live today at $1.36 per million input tokens and $4.18 per million output, and the weights are promised by the end of the month. On Artificial Analysis's DeepSWE 1.1, as charted by Mistral, it scores 62: just ahead of GLM-5.3 at 61, but behind Kimi K3 at 68. In a blind human review of coding quality it placed second, behind only Claude Opus 5. The cyber angle is the real pitch. It scores 82% on reproducing and patching a real vulnerability, the top score on that test, partly because closed models like Claude Opus 5.5 and GPT-6 Astra refuse the task. Mistral says it still refuses malicious cyber prompts more often than any other open model. The fine print: you can't download it yet, the post doesn't name a license, and most charts are Mistral's own picks. Also, 49B active doesn't make it small to host. All 1T parameters still have to fit in memory. Mistral says it beats every open model built in the US or Europe. If the weights land on time, I want to see independent coding numbers before calling it. Source: Mistral's Introducing Mistral Large 4 post, Oct 6, 2026. Independent briefing by CodenamePoshan, not a Mistral post. #MistralAI #OpenSourceAI #LLM #Cybersecurity #AICoding #AIEngineering

About