llmwish.open

IT developer, AI enthusiast, #LLM

[llmwish.open @ciprianveg]

GLM-5.2 on 8× DGX Spark — v18 update 🚀

Jump from 1,200 → 1,329 t/s prefill
Peak decode hits 66 t/s (was 54)
Repo with both v16 and v18 builds:
https://t.co/uFSmpdqw3R

Forum post with benchmark:
https://t.co/mMeqh3Bgcj

[llmwish.open @ciprianveg]

@rodrimora mikrotik crs804-4ddq and 4x4x100 breakout cables

[llmwish.open @ciprianveg]

Preparing 16xGB10 cluster for Kimi K3, Deepseek V4 Pro, and future GLM 5.5 and Minimax M4..

[llmwish.open @ciprianveg]

@HusamSoboh Sigur

[llmwish.open @ciprianveg]

@LD_Theron https://t.co/og26kFbMSP can connect 4, 8 or 16 depending on the breakout cable being used. 400to2x200 or 400to4x100

[llmwish.open @ciprianveg]

Kimi K3 Coder Reap on 8xGB10 - 32t/s avg gen speed in coding task.
This is while waiting for the full 16xgb10 setup to run the full k3. Considering that this one has same no of active experts, it is a good speed for my first, unoptimized, run of KIMI-K3-CODER-REAP-320-MXFP4 with dspark.
I will work on improving these results and let you know.

[llmwish.open @ciprianveg]

@no_stp_on_snek It exists

[llmwish.open @ciprianveg]

@DawsonBotsford 1-1.7kwh

[llmwish.open @ciprianveg]

@u1tra_instinct Awesome!

[llmwish.open @ciprianveg]

@samsaied55 Ray

[llmwish.open @ciprianveg]

@hugoscode Cca 70k

[llmwish.open @ciprianveg]

@tedixh1 Price and better cooling

[keys 🧪 @u1tra_instinct]

🚨🚨🚨🚨🚨: most requested since the DeepSeekV4-Flash GA 0731 release yesterday.

Now Abliterated 32/32 100% compatible with DSpark

please if you like my work and want to contribute please can help via X-money, GoFundMe link in profile 🙏 any little bit helps, whether for token credits, coffee, time, power, hardware.

HF repo:
https://t.co/g5xjS1dOb8

[llmwish.open @ciprianveg]

@0xSero So true..

[llmwish.open @ciprianveg]

@antoniodearmas What is fast enough for you? 50tps+ single stream in coding for glm5.2 is good enough? For me, yes. Multiple requests are also scaling well, till 6-8.

[llmwish.open @ciprianveg]

@antoniodearmas Owning AI instead of being owned by it

[wick @wicksplay]

DeepSeek V4 Flash 0731 @ 118.2 tok/s

First trial. Just getting started. Going to see what I can push this to without sacrificing quality

[llmwish.open @ciprianveg]

@tarat_211 Cca 70k 16 asus, switch and cables

[llmwish.open @ciprianveg]

@tarat_211 6
Cca 70k

[llmwish.open @ciprianveg]

@dev_null321 Old love..

[llmwish.open @ciprianveg]

@defiobi Look at this thread https://t.co/VjZMs91mLC

[llmwish.open @ciprianveg]

@JamesMcPherson CRS804 DDQ with 4x 400to4x100 cables

[llmwish.open @ciprianveg]

@u1tra_instinct AI must be local, even frontier level ones🥂