BasH Rushdn

BasH

[BasH Rushdn @BRushdn]

@antirez Sure why not. Nowadays i find myself keep launching ds4 server anyway for my work, if i can load another lighter model on it, it will be a welcome addition. Perhaps you can throw in some magic as well to optimize it even further like u did with ds4. Cant wait!

[BasH Rushdn @BRushdn]

@HermesAgentTips Qwen3.6 27B ftw! How many vram u got?

[BasH Rushdn @BRushdn]

@konohaart @hqmank I can confirm it works for Malaysia. Not sure about other countries though — you may need a bit of trial and error with the country code (alpha-2 vs alpha-3).

[BasH Rushdn @BRushdn]

2026 is the breakthrough year for running local AI on the Mac.

Thanks to the MLX community and huge leaps in small-to-medium open-weight models like DeepSeek V4, Qwen 3.6, MiniMax 2.7, and Gemma 4 — local AI finally feels practical, capable, and agentic.

[BasH Rushdn @BRushdn]

I bought a maxed out MacBook M4 Max last year hoping to ditch cloud AI subscriptions and cut API costs. But local open-source models were still too far behind in intelligence and agentic capability to be practical.

Come 2026, everything changed.

[BasH Rushdn @BRushdn]

@AlexJonesax @YoussofAl You forgot ds4 by @antirez. Running deepseek 4 flash smoothly on Mac!

[BasH Rushdn @BRushdn]

@bnafOg @jun_song Sounds like AI reply to me 😂