Video wird geladen...

Video konnte nicht geladen werden

Zur Startseite

First kinda working implementation of GLM 5.2 in DwarfStar. Will take some time to be good enough, but it is a promising start. 433 GB GGUF file. M3 Ultra 512GB.

antirez's profile picture

antirez

70,291 subscribers

69,171 Aufrufe • vor 28 Tagen •via X (Twitter)

GamingBildungWissenschaft & Technologie
LIVE

Anya Rossi• Live Now

Private livecam show

0 Kommentare

Keine Kommentare verfügbar

Kommentare vom Original-Post werden hier angezeigt

Ähnliche Videos

I didn't expect DeepSeek v4 PRO (not Flash) to run well on the Mac Studio M3 Ultra with 512GB of RAM. This is 2 bit quantized with the same DwarfStar recipe used for Flash. 433GB GGUF file. 130 t/s prefill, 13 t/s generation. Prefill in the video is low because small prompt.
1:24

I didn't expect DeepSeek v4 PRO (not Flash) to run well on the Mac Studio M3 Ultra with 512GB of RAM. This is 2 bit quantized with the same DwarfStar recipe used for Flash. 433GB GGUF file. 130 t/s prefill, 13 t/s generation. Prefill in the video is low because small prompt.

antirez

169,897 Aufrufe • vor 2 Monaten

I did it! It works! Using GLM-4.7-4bit with mlx_lm.server and opencode to fix real code locally! 🔥 Here single M3 Ultra 512GB, nex step phase will be 2 using Tensor Parallelism and then apply same changes to exo. Prefill is slow on a single machine, but generation is good.
1:44

I did it! It works! Using GLM-4.7-4bit with mlx_lm.server and opencode to fix real code locally! 🔥 Here single M3 Ultra 512GB, nex step phase will be 2 using Tensor Parallelism and then apply same changes to exo. Prefill is slow on a single machine, but generation is good.

Ivan Fioravanti ᯅ

44,000 Aufrufe • vor 6 Monaten

MLX GLM 5.2 Distributed on two M3 Ultra 512GB 🔥 One M3 Ultra: 18.8 tokens/sec Two M3 Ultra: 23.4 tokens/sec Context: - PR by Pedro Cuenca is still open and probably there is room for improvement: - basic generation test to measure decoding performance here, I will do a full context benchmarking once PR is more mature - nvfp4 quantization used - Video alternates standard speed and x20, with one Mac first and distributed later. Enjoy! 🙌🏻
0:35

MLX GLM 5.2 Distributed on two M3 Ultra 512GB 🔥 One M3 Ultra: 18.8 tokens/sec Two M3 Ultra: 23.4 tokens/sec Context: - PR by Pedro Cuenca is still open and probably there is room for improvement: - basic generation test to measure decoding performance here, I will do a full context benchmarking once PR is more mature - nvfp4 quantization used - Video alternates standard speed and x20, with one Mac first and distributed later. Enjoy! 🙌🏻

Ivan Fioravanti ᯅ

87,805 Aufrufe • vor 28 Tagen

GLM-5 runs with mlx-lm on a single 512GB M3 Ultra in Q4. It's quite good in my initial testing and pretty fast as well. It generated a highly functional space invaders game using 7.1k tokens at 15.4 tok/s and 419GB memory. Thanks to Gökdeniz Gülmez and Tarjei Mandt for the port.
1:11

GLM-5 runs with mlx-lm on a single 512GB M3 Ultra in Q4. It's quite good in my initial testing and pretty fast as well. It generated a highly functional space invaders game using 7.1k tokens at 15.4 tok/s and 419GB memory. Thanks to Gökdeniz Gülmez and Tarjei Mandt for the port.

Awni Hannun

60,599 Aufrufe • vor 5 Monaten

MPJ: "Some of the shots and some of the ways they (the Nuggets) probably could have used me we had enough good players to where we could create a good shot in a different way...it is good to be able to kinda show what I'm capable of"
1:30

MPJ: "Some of the shots and some of the ways they (the Nuggets) probably could have used me we had enough good players to where we could create a good shot in a different way...it is good to be able to kinda show what I'm capable of"

Oh No He Didn't

158,270 Aufrufe • vor 6 Monaten

GLM 5.2, Q2_K routed experts (effectively ~2.6 bits) running with SSD streaming on an M5 Max 128GB computer.
1:31

GLM 5.2, Q2_K routed experts (effectively ~2.6 bits) running with SSD streaming on an M5 Max 128GB computer.

antirez

133,452 Aufrufe • vor 21 Tagen

DS4 running on DGX Spark (GB10 / CUDA), private branch for now. 12 tokens/sec, the memory bandwidth is limited in this system, at 270GB/sec. But prefill is ways more alighed to M3 Max at ~200 t/s. I'll release when more mature, but it is almost sure that it will get merged.
1:34

DS4 running on DGX Spark (GB10 / CUDA), private branch for now. 12 tokens/sec, the memory bandwidth is limited in this system, at 270GB/sec. But prefill is ways more alighed to M3 Max at ~200 t/s. I'll release when more mature, but it is almost sure that it will get merged.

antirez

84,057 Aufrufe • vor 2 Monaten

Conservative prosecutors are resigning instead of carrying out Trump DOJ political orders. “You will eventually find someone who is enough of a fool, or enough of a coward, to file your motion. But it was never going to be me.” Republicans in Congress—STAND UP.
1:42

Conservative prosecutors are resigning instead of carrying out Trump DOJ political orders. “You will eventually find someone who is enough of a fool, or enough of a coward, to file your motion. But it was never going to be me.” Republicans in Congress—STAND UP.

Amy Klobuchar

385,534 Aufrufe • vor 1 Jahr

Zavier Gozo is the best attacking talent I’ve seen out of MLS in quite some time. Believe the hype; he’s good enough to start in a Champions League team.
1:01

Zavier Gozo is the best attacking talent I’ve seen out of MLS in quite some time. Believe the hype; he’s good enough to start in a Champions League team.

Zach Lowy

86,496 Aufrufe • vor 2 Monaten

Ran two prompts in Kimi K2.7 Code HighSpeed, Kimi K2.7 Code, and GLM 5.2 HighSpeed is 2.5 to 4.2 times faster than Kimi K2.7 Code while producing comparable quality 4.5 to 15 times faster than GLM 5.2, and half the price (or less) These are simple tests, so the pricing difference is small and speed difference could diverge a lot But definitely a hell of a lot faster with good output
0:53

Ran two prompts in Kimi K2.7 Code HighSpeed, Kimi K2.7 Code, and GLM 5.2 HighSpeed is 2.5 to 4.2 times faster than Kimi K2.7 Code while producing comparable quality 4.5 to 15 times faster than GLM 5.2, and half the price (or less) These are simple tests, so the pricing difference is small and speed difference could diverge a lot But definitely a hell of a lot faster with good output

GMI Cloud

38,008 Aufrufe • vor 25 Tagen

Much better at 17 tokens/sec in M3 Max, fusing a few operations. 128gb m3 max.
0:49

Much better at 17 tokens/sec in M3 Max, fusing a few operations. 128gb m3 max.

antirez

25,147 Aufrufe • vor 2 Monaten

“I expect you will eventually find someone who is enough of a fool, or enough of a coward, to file your motion. But it was never going to be me.” A career prosecutor stood up to Emil Bove’s lawlessness. Republicans in the Senate should do the same and reject his nomination.
1:24

“I expect you will eventually find someone who is enough of a fool, or enough of a coward, to file your motion. But it was never going to be me.” A career prosecutor stood up to Emil Bove’s lawlessness. Republicans in the Senate should do the same and reject his nomination.

Amy Klobuchar

261,408 Aufrufe • vor 1 Jahr

Bryce Lance - WR - NDSU If GB is going to take a WR. He is going to be very packery. (Would guess he will be a tier 1 fit) He never got to play with Christian Watson at NDSU. But maybe he can in GB.
0:20

Bryce Lance - WR - NDSU If GB is going to take a WR. He is going to be very packery. (Would guess he will be a tier 1 fit) He never got to play with Christian Watson at NDSU. But maybe he can in GB.

Jacob Morley

100,354 Aufrufe • vor 5 Monaten

I was a bit suspicious of the claim, but GLM-5.2 is pretty good at designing stuff. Obviously not at the level of a professional designer, but it has that Opus-level quality. Great at: - games - landing pages - HTML artifacts - 3D worlds Wish I had Fable 5 to compare with.
9:24

I was a bit suspicious of the claim, but GLM-5.2 is pretty good at designing stuff. Obviously not at the level of a professional designer, but it has that Opus-level quality. Great at: - games - landing pages - HTML artifacts - 3D worlds Wish I had Fable 5 to compare with.

elvis

18,938 Aufrufe • vor 1 Monat

GLM 5.2 is the best Open Model in the world, and many are saying it's almost as good as Opus 4.8 at a fraction of the cost. Here's exactly how to set it up in Cursor using OpenRouter. Complete instructions below 👇
1:58

GLM 5.2 is the best Open Model in the world, and many are saying it's almost as good as Opus 4.8 at a fraction of the cost. Here's exactly how to set it up in Cursor using OpenRouter. Complete instructions below 👇

Agent Native

20,658 Aufrufe • vor 29 Tagen

This was the start of Messi hatred for a lot of Nigerians, but some of us were objective enough to see how Good Messi truly is and moved on
1:24

This was the start of Messi hatred for a lot of Nigerians, but some of us were objective enough to see how Good Messi truly is and moved on

Leonard Onochie

132,527 Aufrufe • vor 1 Monat

Discord is working on voice filters that will allow you to change your voice in VCs. Some of these filters will only be available to Discord Nitro subscribers, but it is not yet clear which ones.
0:23

Discord is working on voice filters that will allow you to change your voice in VCs. Some of these filters will only be available to Discord Nitro subscribers, but it is not yet clear which ones.

Discord Previews

86,665 Aufrufe • vor 1 Jahr

Very disappointed that the CX will not give a date when the Defence Investment Plan will be published despite the government first promising it in June last year … it will lead many to have grave concerns about the commitment to investment in defence spending at a time of crisis.
0:42

Very disappointed that the CX will not give a date when the Defence Investment Plan will be published despite the government first promising it in June last year … it will lead many to have grave concerns about the commitment to investment in defence spending at a time of crisis.

Rt Hon John Glen MP

10,368 Aufrufe • vor 4 Monaten