DeepCoder 14B LOCAL Test & Install (A THINKING Coding Model) скачать в хорошем качестве

DeepCoder 14B LOCAL Test & Install (A THINKING Coding Model)

Timestamps: 00:00 - Intro 02:00 - LMStudio Install 03:54 - Python Game Test 06:00 - Second Game Test 10:10 - Web Dev Test 13:02 - Refusal Test 14:07 - Closing Thoughts In this video, we test DeepCoder-14B-Preview, a new code-focused LLM fine-tuned from DeepSeek-R1-Distilled-Qwen-14B, and designed specifically for reasoning and programming tasks. What makes this model especially interesting is that it can be run locally on a single 12GB RTX 3060 GPU using the Q4_K_M quantized GGUF, making it accessible to far more users. We start with a quick overview of the model and install it using LMStudio. From there, we run a series of tests to evaluate its coding performance. We begin by prompting it to build a Python retro-style game, followed by a second test where we try to push for a more polished result. Next, we ask it to create a basic HTML web page, which results in a clean and fully functional demo. Finally, we test the model's boundaries with a refusal prompt, probing how it handles security-related requests. HF Repo: https://huggingface.co/agentica-org/D...