Qwen3.6 Tackles Tough Reverse Engineering Challenge π±
Watch Qwen3.6 35B conquer a tough reverse engineering task, building an LTE modem crawler from scratch and matching Claude Sonnet.

Protorikis
38.0K views β’ Apr 28, 2026

About this video
In this video, I put Qwen3.6 35B A3B through the ultimate real-world coding challenge: building a working LTE modem crawler from scratch. After previous model Gemma 4 failed, this task became the perfect benchmark for memory, reasoning, and persistence. It tests context handling, reverse engineering, and navigating unknown authentication flows. Across two days we push this local LLM to its limits to find pure amazement.
You'll learn:
- How powerful local LLMs have become on complex real-world coding tasks
- How to select the right model for the task based on custom benchmarks
- How Qwen3.6 handles long-context reasoning compared to Gemma 4
- Practical techniques for working with local LLMs on large codebases
- How a local model closes-in on Claude Sonnet
- Performance insights: prefill vs decode speed in large contexts
I chose Qwen3.6 based on the 6 custom-built bechmarks done previously.
The 3 episodes that lead to this video (in watch order):
π https://youtu.be/cBoWEQVWUVs - Gemma 4 vs Qwen: Easy Wins... and One Critical Weakness
π https://youtu.be/ONQcX9s6_co - Qwen3.6 vs Gemma 4: Which Actually Remembers Your Code?
π https://youtu.be/In825VzHzbU - Qwen3.6 27B vs Gemma 4 31B: Memory Recall Battle with a Single Winner
The KV cache video (performance insight):
π https://youtu.be/QIZz4AF0U24
Models tested:
* Qwen3.6 35B A3B Q4 K M lmstudio-community
Models referenced:
* Gemma 4 26B A4B Q4 K M
* Gemma 4 31B Q4 K M
* Qwen3.5
Hardware for reference:
* MacBook Pro M3 Max 36GB
π€ Business inquiries: contact@protorikis.com
β±οΈ Chapters
00:00 - Intro (The 3 Episodes that Lead to This)
01:08 - Gemma 4βs Sliding Window Attention
01:26 - LTE Modem Crawler Challenge Explained
01:59 - Large Context Hallucinations and Memory Recall Benchmark
02:54 - The Coding Challenge Setup
03:20 - Download Client-Side Files
04:40 - Beautify Downloaded JavaScript Files
05:07 - Reverse Engineer Login Flow
06:27 - The Missing Files
08:05 - Get Radio Signal Metrics
09:12 - The Source (Is Within You)
09:38 - Qwen3.6 vs Claude Sonnet 4.6
10:00 - Performance
10:28 - Conclusion
You'll learn:
- How powerful local LLMs have become on complex real-world coding tasks
- How to select the right model for the task based on custom benchmarks
- How Qwen3.6 handles long-context reasoning compared to Gemma 4
- Practical techniques for working with local LLMs on large codebases
- How a local model closes-in on Claude Sonnet
- Performance insights: prefill vs decode speed in large contexts
I chose Qwen3.6 based on the 6 custom-built bechmarks done previously.
The 3 episodes that lead to this video (in watch order):
π https://youtu.be/cBoWEQVWUVs - Gemma 4 vs Qwen: Easy Wins... and One Critical Weakness
π https://youtu.be/ONQcX9s6_co - Qwen3.6 vs Gemma 4: Which Actually Remembers Your Code?
π https://youtu.be/In825VzHzbU - Qwen3.6 27B vs Gemma 4 31B: Memory Recall Battle with a Single Winner
The KV cache video (performance insight):
π https://youtu.be/QIZz4AF0U24
Models tested:
* Qwen3.6 35B A3B Q4 K M lmstudio-community
Models referenced:
* Gemma 4 26B A4B Q4 K M
* Gemma 4 31B Q4 K M
* Qwen3.5
Hardware for reference:
* MacBook Pro M3 Max 36GB
π€ Business inquiries: contact@protorikis.com
β±οΈ Chapters
00:00 - Intro (The 3 Episodes that Lead to This)
01:08 - Gemma 4βs Sliding Window Attention
01:26 - LTE Modem Crawler Challenge Explained
01:59 - Large Context Hallucinations and Memory Recall Benchmark
02:54 - The Coding Challenge Setup
03:20 - Download Client-Side Files
04:40 - Beautify Downloaded JavaScript Files
05:07 - Reverse Engineer Login Flow
06:27 - The Missing Files
08:05 - Get Radio Signal Metrics
09:12 - The Source (Is Within You)
09:38 - Qwen3.6 vs Claude Sonnet 4.6
10:00 - Performance
10:28 - Conclusion
Tags and Topics
Browse our collection to discover more content in these categories.
Video Information
Views
38.0K
Likes
1.7K
Duration
10:51
Published
Apr 28, 2026
User Reviews
4.7
(7) Related Trending Topics
LIVE TRENDSRelated trending topics. Click any trend to explore more videos.