After weeks of work, we have finally brought our custom DeepSeek V4.1 Flash onto our local cluster
Built on two Strix Halo AI Max+ systems with a total of 256GB unified memory and interconnected through a Mellanox 100GbE network. At the same time, we completed a LoRA fine
Top-5 Best Value #LLM Models of the Day at UTC-04
| Model | #AAII | Price | | GLM 5.3 Flash | 42 | $0.05 | | DeepSeek V4.1 Flash | 40 | $0.08 | | GLM 5.3 | 45 | $0.42 | | GPT-6 Luna | 37 | $0.08 | | MiMo-V2.6-Pro | 46 | $0.32 |
Astra vs Liang: can DeepSeek hit a 10-month payback on Huawei Ascend?
Accelerator cost is the whole game. Stack ↓
In the past two days, I saw that everyone is using Opus 5.5 to directly generate videos.
In the past two days, I saw that everyone is using Opus 5.5 to directly generate videos. Since my account was blocked, I tried DeepSeek Harness Desktop + deepseek-flash. The DeepSeek Harness Desktop version I am using is 0.1.7-rc.2 and the model is deepseek-flash. 1. The prompt words are directly used by Teacher Baoyu: --- Prompt word ---…
You have reached the end of the archive
All of deepseek