TwIL LM3 Pro: Logic Model Locally on Contradictions and Reasoning Puzzles @fahdmirza
TwIL LM3 Pro: Logic Model Locally on Contradictions and Reasoning Puzzles  @fahdmirza
Uploaded November 2026 | Updated November 2026, 4 days ago
This video installs and tests TwIL-LM3-Pro which is a 3.66B reasoning model for formal logic tasks.

📬Weekly AI Newsletter: fahdmirza.substack.com

🔥 Get 50% Discount on any A6000 or A5000 GPU rental, use following link and coupon:

bit.ly/fahd-mirza
Coupon code: FahdMirza

🔥 Buy Me a Coffee to support the channel: ko-fi.com/fahdmirza

#TwILLM3Pro

PLEASE FOLLOW ME:
â–¶ LinkedIn: / fahdmirza
â–¶ YouTube: / @fahdmirza
â–¶ Blog: fahdmirza.com

00:00 Intro: what TwIL-LM3-Pro is
0:38 Test 1: Policy checker
02:20 Test 2: Hidden contradictions in a festival ticketing spec
04:23 Test 3: Invented chemistry rules
05:04 Benchmark chart
06:20 Test 4: Masked ball puzzle
08:07 How it was trained
08:55 Verdict and where to use it

RESOURCES:

â–¶ huggingface.co/Nanbeige/Nanbeige4.3-33B

All rights reserved © Fahd Mirza
TwIL LM3 Pro: Logic Model Locally on Contradictions and Reasoning PuzzlesFlint 40B by Recursive: Fast Local Coding AI TestedVoice of Reason from Kyutai on Top of GLM: Run LocallyRun Ornith-1.5 122B Locally: Stronger, Faster, LeanerJev: The Model That Killed Chat GPTs Core Idea? RLCD ExplainedPhonon-2: 164MB Accurate Speech Recognition Model: Run LocallyStop Using OCR - EVIE: The #1 Visual Document Retriever - Run LocallyPG-Jev: Query Your Database in Plain English with Decision ModelKimi K3.1 Pro: Most Capable Open-Weight Model NowPixel Canary: The Free Stealth Model Beating Expensive ModelsGemini 3.8 Flash: Google is Back on AI Horse: Cheap and FastTernary Bonsai 2 9B in GGUF: A 2.4 GB Model Locally
Fahd Mirza |

TwIL LM3 Pro: Logic Model Locally on Contradictions and Reasoning Puzzles

SHARE TO X SHARE TO REDDIT SHARE TO FACEBOOK WALLPAPER