Asim Abdul Ghafoor

Verified account

@AsimAbdulG

AI and hardware

Pakistan

Joined July 2009

219 Following

27 Followers

125 Posts

Asim Abdul Ghafoor

about 12 hours ago

Today I tried Strata on my Mac M5 Max 128GB with Qwen Flash Next unsloth-ud-q4_k_xl. Truly amazing performance, constant 89 Tok/s. I even selected to leave room for other work (Not at full performance). Well done! @coldniko Thank you! I have switched from mlx-serve.

1

1

0

0

61

Asim Abdul Ghafoor

2 days ago

@9to5mac @apollozac I use JumpDesktop. It is superior product and almost free. Only one time small Fee.

0

0

0

0

2

Asim Abdul Ghafoor

2 days ago

@Beamsters1 I understand. I am trying Qwen4 Max in Qwen Cloud. Lets see how it compares to Opus

1

0

0

0

50

Asim Abdul Ghafoor

2 days ago

@Beamsters1 How can I help

0

0

0

0

5

Who to follow

I M Here not to change the world but to be myself!!!!

@Rizwan_Ghafoor

BSc, MBBS, MRCPS(Glasg), FCPS(Tr&Orth), FACS(USA) | Member, American Academy of Orthopaedic Surgeons | Consultant Trauma & Orthopaedic Surgeon

@naeemwyneofficial

Social and digital media strategist with more than two decades of experience across broadcast media, public institutions and creative environments.

Asim Abdul Ghafoor

2 days ago

@Beamsters1 Please, please bring ernie-5.5-flash to Sushi

2

0

0

0

19

Asim Abdul Ghafoor

2 days ago

@Beamsters1 I hope so. I heard 11.11 for Qwen4 (Q-Waiting)

0

1

0

0

61

Asim Abdul Ghafoor

3 days ago

@Alibaba_Qwen Awaiting begins.....

0

0

0

0

166

Asim Abdul Ghafoor

3 days ago

@coffeecup2020 They should not have given a teaser when it will be launched.

0

1

0

0

353

Asim Abdul Ghafoor

3 days ago

Running deepseek 4.1 STQ using Sushi 1.6.0 on M5 Max 128GB under compressed air cooler to lower GPU temperatures. Removed seals for macbook @Beamsters1

AsimAbdulG's tweet photo. Running deepseek 4.1 STQ using Sushi 1.6.0 on M5 Max 128GB under compressed air cooler to lower GPU temperatures. Removed seals for macbook @Beamsters1 https://t.co/hnc5NWLqws

1

6

0

2

182

Asim Abdul Ghafoor

4 days ago

@Beamsters1 Looking to build this for cooling: https://t.co/2gmVgmIGJD

0

3

0

2

562

Asim Abdul Ghafoor

4 days ago

@Beamsters1 100% Agree, I need some good water cooled body underneath it. Right now I have 6 Fans cooler

1

3

0

0

1K

Asim Abdul Ghafoor

4 days ago

@kimmonismus If you have got 128GB of RAM and M5 Max, go for sushi 1.5.1 and Deepseek STQ 1.75bwp. Twice as better than Qwen3.8-Flash-Next. His work @Beamsters1

0

1

0

0

53

Asim Abdul Ghafoor

4 days ago

Thank you @Beamsters1 After much testing and effort to free up as much memory on M5 MAX 128GB I have, finally switched to sushi 1.5.1 and DeepSeek STQ 1.75bpw. What an amazing model, work much better than Qwen3.8-Flash-Next. WOW! Now I need some water cooled body to put underneath my macbook max. Turned off Siri as well to save memory.

2

2

0

0

147

Asim Abdul Ghafoor

5 days ago

@Rizwan_Ghafoor @StormFiberPK Yes, I agree. Recently stormfiber in Multan is slow.

1

0

0

0

64

Asim Abdul Ghafoor

5 days ago

@RuiCarrilho5 Start using deepseek if you have 128gb Ram and M5 Max. It is more capable model. Use sushi 1.5.1

0

0

0

0

80

Asim Abdul Ghafoor

7 days ago

@Alibaba_Qwen Eagarly waiting for it....

0

1

0

0

98

Asim Abdul Ghafoor

12 days ago

Please note model quality for Qwen3.8-Flash-Next matters for large / complex projects on MLX Enginees. Right now best one is: ddalcu/Qwen3.8-Flash-Next-MLX-Serve-iQ-MLX-4.7bpw Excelletn quality/speed on M5 Max 128GB

1

1

0

0

54

Asim Abdul Ghafoor

12 days ago

I have tested lots of Local AI MLX Engines for M5 Max 128GB. Below are top ones in order: Model: ddalcu/Qwen3.8-Flash-Next-MLX-Serve-iQ-MLX-4.7bpw 1. Mlx-Serve (Recommended) 2. Sushi 3. OMLX 4. Strata 5. MTPLX 6. LMStudio 7. Ollama Waiting EXL....

1

3

0

0

181

Last Seen Users on Sotwe

Trends for you

Most Popular Users