
We didn’t find a product being promoted in this video. Its hook, format and structure below still show why it took off.
Browse products winning on TikTok1. What it is: a 24B open-weight reasoning model that only wakes about 2B parameters per token. Each token goes to 4 of 64 experts. 131K context, trained on 33T tokens. 2. Fast on a laptop: 184 tokens/s on an Apple M5 Max and 114 on an AMD Ryzen AI Max+ 395, up to 1.7x faster than comparable MoEs on-device, Liquid AI says. Over 21K tokens/s on one NVIDIA B300. 3. Where it leads: AutomationBench 27.69 vs 8.72 for Qwen3.6-35B-A3B, plus ToolSandbox, LiveClaw, IFBench and IFEval. 4. Where it trails: SWE-Bench Verified 51.0 vs 66.2 and GPQA Diamond 74.4 vs 83.7 for Qwen3.6. Liquid AI says to pair it with retrieval or web search for knowledge-heavy work. The catch: these are Liquid AI's own numbers, it always reasons before answering, and it's text-only. Follow @codenameposhan for more AI breakdowns. Sources: Liquid AI's launch blog and the Hugging Face model card (Oct 13, 2026). Independent briefing by CodenamePoshan, not a Liquid AI post. Liquid AI logo and launch art are © Liquid AI; benchmark values are from Liquid AI, shown for editorial reporting. #LiquidAI #LFM #OnDeviceAI #AI #AIAgents #LLM
Promotes 7 products. Create free account to see more.
Outliers and viral hits: their videos on Overshot, all time.