I ran two tests on my M5 Max w/ 128GB inside DeepSeek Harness.
On the left is DwarfStar 4 with the 4bit quant. It averaged 52tok/s and took nearly 100 minutes to complete. On the right is MLX-Serve with the 4bit/8bit quant. It averaged 72tok/s and finished in 50 minutes. Both
Recently, I used AI to make a very practical little thing. I used DeepSeek 4.1 Flash to make a small program called "Don't Forget When You Go Out".
The logic is simple: Before going out, fill in the place of departure, destination, departure and return times, means of transportation, and number of people traveling with you, and it will automatically generate a list of items to carry based on the trip.
Five frontier models. One simple prompt.
No retries. No blender. Just a simple frontend web design. Fable 5.1 vs. GPT-6 Astra vs. DeepSeek v4.1 Flash vs. GLM 5.3 Flash vs. Qwen3.8 Flash Next Which model built the best looking luxury watch landing page? IMO Fable 5.1 and GLM
LOCALS ONLY ! deepseek x nvidia .
Have a good day ya'll
You have reached the end of the archive
All of deepseek