thomas lanning

[Lamb Labs (YC S26) @LambLabs]

Meet Woolly 🐑

We post-trained Qwen3-8B to run up to 2–3× faster on math & coding prompts and, more importantly, to fit on our chip.

The technique works with any LLM, regardless of size or quantization.

Live demo this week → https://t.co/7K5HUfVq5Z

[thomas lanning @thomasslanning]

@plugyawn Think we can help. Can’t dm you though

[Sam @Sam8002642]

Pov : 2 hours before deadline

[thomas lanning @thomasslanning]

@Rivermaan45 @JasperLort I’ll buy them all!🐑🐑

[thomas lanning @thomasslanning]

@HaseebRaza715 @LambLabs To prove no animals were harmed in the making of this video ;)

[thomas lanning @thomasslanning]

@aryaan_smm @LambLabs Lot of energy now in inference rn is going towards pulling weights from memory, and there’s quite a few tricks to reduce that. Both hardware and software

[thomas lanning @thomasslanning]

@LambLabs stay tuned next week for another launch 🐑 (on the more software side)

[Lamb Labs (YC S26) @LambLabs]

Lamb Labs (YC S26) is building custom AI inference chips that can do 20,000+ tok/s at 63x higher Intelligence per Watt (IPW) than traditional GPUs.

GPUs became the default for AI because they were available, not because they were the right hardware.