Qwen 3.6 14B A3B FableVibes benchmarked and tested vs Base Qwen 35B - 16GB Local LLM setup
In this video I want to check out Qwen 3.6 14B A3B FableVibes from tvall43. Starting with Qwen3.6-35B-A3B-heretic the model was then pruned down to 14B via REAP. Let's see how the speed boost affects intelligence. I run through a few coding tests which are: 1. Performance 2. Memory 3. Agency 4. OpenAI Human Eval 5. Sand Physics 6. Dungeon Crawler If you're interested in local LLMs, AI and homelabs - feel free to subscribe! Fablevibes: https://huggingface.co/tvall43/Qwen3.... Qwen: https://huggingface.co/unsloth/Qwen3.... GitHub: https://github.com/lukesdevlab/youtube HumanEval: https://github.com/openai/human-eval Patreon: / lukesdevlab #localllm #localai #homelab #llamacpp #homelab #openai #qwen #35B #14B #fablevibes #A3B Chapters: 0:00 Intro 0:14 Model 1:07 Tests Overview 2:09 Target System Specs 2:25 Performance 3:49 Memory 4:52 Agency 7:27 OpenAI HumanEval 8:53 Sand Physics - 14B Q4 11:55 Sand Physics - 14B Q8 13:38 Sand Physics - Qwen 35B 14:46 Dungeon Crawler - 14B Q4 16:45 Dungeon Crawler - 14B Q8 18:21 Dungeon Crawler - Qwen 35B Q4_K_M 19:52 Dungeon Crawler - Qwen 35B IQ4_NL 21:13 Conclusion
This Is What the Start of a Crash Looks Like
Ternary Bonsai 27B benchmarked and tested vs Qwen 27B - 16GB Local LLM setup
The Most Important Conversation in AI Right Now
OpenAI's Bots Break Containment and Hack Hugging Face Autonomously — With Alex Stamos
Harnesses in AI: A Deep Dive — Tejas Kumar, IBM
Gemini 3.6 Flash Is HERE – Testing Google’s BEST Model Yet!
Running a local LLM with two GPUs
Ornith 35B Benchmarked vs Qwen 35B - 16GB Local LLM setup
Gemma 4 26B A4B QAT vs non-QAT - 16GB Local LLM setup
Scott Ritter: ‘Turn Around or Face God’: Yemen Issues Chilling Warning to Saudi Ship in the Red Sea
I Built an LLM From Scratch
Kimi K3 explained in 13min..
You Can Learn AI Agent Harness & Loop Engineering In 19 Min | LLM Ops, Eval, Tracing, RAG
Laguna S 2.1 The BEST LOCAL Model? Open-Weight Model Beats GLM 5.2? (FULLY FREE)
Kimi 3: AI still isn’t profitable | David Gerard
No Boss, No Money: The Raw Reality of China’s Gen-Z Freelancers
Little Coder: Small Models Need Small Harnesses
Agents A1 Benchmarked vs Qwen 35B - 16GB Local LLM setup
China schockt alle: Kimi K3 ist so gut, dass die Server kapitulieren
