gelim

Mixing colors in cybersec

[gelim @gelim]

After reading @Guardicore CVE-2020-3952 excellent article and PoC I rewrote it a bit to check without adding admin user here https://t.co/u2fxW8flr5
#vmware #vmdir #ldap #vcenter #vulnerability

[gelim @gelim]

@nullfoundry If you want to find the best batch and ubatch for your card use llama-bench with several values to check the best throughput

[gelim @gelim]

@0xSero IMHO not worth the vram gap when you can run qwen3.6 27b full context in 32 GB VRAM

[gelim @gelim]

@GPTWare Qwen3.7 27B is a blast. You should try it

[gelim @gelim]

@lexfridman @FFmpeg Thanks for this awesome release Lex. I feel that @ggml_org #llamacpp is the AI era equivalent to @videolan / @FFmpeg

[gelim @gelim]

@LottoLabs Qwen3.7-35B-A3B with Qwen3.6-27B super power :-)

[gelim @gelim]

@ggerganov With Pi harness. Detail not left unnoticed :)

[gelim @gelim]

@The_Only_Signal Thx this confirm my expectation : not worth the additional compute / vram if you can get out with qwen3.6 27B and maybe steer it a bit more.

[gelim @gelim]

@antirez Maybe worth conparing with Minimax M2.7 UD-IQ4_XS or some of the Q3?

[gelim @gelim]

@Tur24Tur @deepseek_ai @PortSwigger @claudeai Curious how qwen3.6-27B would perform on those tests

[gelim @gelim]

@MiniMax_AI @vllm_project It seems @UnslothAI + @ggml_org gives us llama.cpp day 0 support as well

[gelim @gelim]

@victormustar brace for the incoming wave of download :-)

[gelim @gelim]

@ivanfioravanti https://t.co/OnnPIu8R6x

[gelim @gelim]

@NielsRogge How was the speed in minimax? (pp & tg) and how much context max at which quant? Using llama.cpp I guess if speaking about Q3

[gelim @gelim]

@ivanfioravanti @nyan4maru @grok pointless to stick on DGX Spark that is a bit overpriced. I'll take 2 Asus Ascent GX10 for the price of your MBP M5 Max. 2x3k€ vs. 6k€. Same NVme storage in total, twice the unified RAM. What about thermal throttling on sustained load? 🍊 & 🍏:-)

[gelim @gelim]

@Carlos_Ajoy @TeksEdge As to how much the quality is impacted with this quantization level... That stays unanswered 😕

[gelim @gelim]

@Carlos_Ajoy @TeksEdge UD-Q3_K_XL (unsloth GGUF) and max 64k context in Q8. Probably better to have the machine headless to spare some RAM.

[gelim @gelim]

@SkylerMiao7 Goat. ♥️

[gelim @gelim]

https://t.co/6uhicxNmhk

[Skyler Miao @SkylerMiao7]

M2.7 open weights coming in ~2 weeks.
still actively iterating
just updated a new version on yesterday — noticeably better on OpenClaw.

[gelim @gelim]

@MiniMax_AI M2.7 ♥️🤗?

[gelim @gelim]

@iamsupersocks Breaking news qui fait bien plaisir https://t.co/6uhicxNmhk

[Skyler Miao @SkylerMiao7]

M2.7 open weights coming in ~2 weeks.
still actively iterating
just updated a new version on yesterday — noticeably better on OpenClaw.