Umer Farooq @Dev__F

AI consultant • Building AI that actually works! https://t.co/ctLLJKKaZs umerfarooq.me Joined August 2023

Tweets

1K
Followers

146
Following

523
Likes

97

GMI Cloud @gmi_cloud

23 hours ago

DeepSeek V4 Pro is now 61% off on GMI for unlimited time $0.678 per 1M token input $1.357 per 1M token output available on open router and via GMI api key 👇

5 14 213 20K 48

View Details

Umer Farooq @Dev__F

2 days ago

@minoqtopus Asli PakWheels “humari koi or branch nhi hai” xD

0 0 0 39 0

View Details

Umer Farooq @Dev__F

4 days ago

@boardyai Commenting so yoy dont have to find

0 0 0 12 0

View Details

Qwen3.8 is launching and going open-weight soon!🌐 With a massive 2.4T parameters, this model is continuously evolving. We believe it’s one of the most powerful model available today, compatible to leading frontier AI models , second only to Fable 5. You don't have to wait to

1K 3K 24K 8.2M 5K

0 0 0 31 0

View Details

Umer Farooq @Dev__F

6 days ago

@thsottiaux Codex usage reset reminder for @thsottiaux ❤️.Need some support

0 0 0 60 0

View Details

Umer Farooq @Dev__F

a week ago

@bakkermichiel I don’t believe in benchmarks, they can be you can say rigged. I will test it on my daily tasks and then will leave feedback

0 0 3 723 0

View Details

Umer Farooq @Dev__F

a week ago

Let’s gooo

GMI Cloud @gmi_cloud

a week ago

Kimi K3 is available to all GMI users 30% off, 48 hrs, unlimited Go build 👇

2 4 23 2K 1

0 0 0 59 0

View Details

Umer Farooq @Dev__F

a week ago

@thsottiaux Let’s celebrate with a reset 😍

0 0 2 262 0

View Details

Umer Farooq @Dev__F

a week ago

@thsottiaux Instead of merging, I would suggest dissolving something 👀 #iykyk

0 0 0 479 0

View Details

Umer Farooq @Dev__F

a week ago

Source: huggingnews.com/ai/moonshot-la…

0 0 0 15 0

View Details

Umer Farooq @Dev__F

a week ago

2.8 trillion parameters sounds like the headline. It isn't. Moonshot just launched Kimi K3 on the web and through its API. 1 million-token context. A score of 1687 on GDPval-AA v2, above Claude Opus 4.8 in Moonshot's benchmark materials and below GPT-5.6 Sol at 1747.8. And open weights are supposed to land within days. That last part matters more than the giant parameter count. I build AI systems for a living. Most clients don't care which model owns a benchmark for two weeks. They care whether they can choose the provider, control the data path, fine-tune for their workflow, and avoid rebuilding everything when one vendor changes pricing. If K3's weights actually land and the model holds up outside vendor benchmarks, this isn't just another big Chinese model release. It's another chunk taken out of the closed-model moat. 2.8T gets the attention. Open weights change the buying decision. Would you put K3 into a production eval now, or wait until the weights and independent benchmarks are out? #AI #LLM #OpenSourceAI #AIAgents

3 0 0 111 0

View Details

Umer Farooq @Dev__F

a week ago

Worth noting: the 1687 score comes from Moonshot's own benchmark materials. I'm waiting for independent testing before treating the ranking as settled.