Fl_r

Agibot A3 just unlocked Webster flips, Thomas flares, and mid-air running flips 🤸‍♂️🤖 At 173 cm and 55 kg, this full-size humanoid turns a 0.218 kW/kg power-to-weight ratio into continuous, fluid, gymnastic-grade motion control.

MTP speedup Qwen by 2.5x in Atomic Chat Dense vs MoE models on 2x RTX 5090 Qwen3.6 27B: 51 → 117 tps +137% Qwen3.6 35B-A3B: 218 → 267 tps +25% MTP drafts several tokens ahead and verifies them in one pass. The speedup depends on memory moved per pass. Dense 27B reads all 27B params per token, MoE 35B-A3B only reads 3B active. Dense had way more to save by batching. The baseline tps also differ (218 vs 51) for the same reason from the other side. Token generation is memory-bandwidth bound, and MoE moves ~8x less memory per token, so its baseline is already 4x ahead. ~80% draft acceptance. Zero accuracy loss. ~1 GB extra VRAM. Open-source code and local AI app – in the comments 👇

We just wrapped what began as an 8-hour challenge - and it ran for 200 hours without a failure Shoutout to the team for the hardcore engineering behind F.03 and the robust Helix models powering it