model roundup

DeepSeek 4

3 items · started 2026-07-23 · closed 2026-07-26

We self-host DeepSeek V4 Flash on AWS spot instances (twitter.com via hn)

+1 8w deepseek

https://t.co/syBl6fkTDJ Miguel Salinas@VercantezHow we self-host DeepSeek V4 Flash on AWS spot instances4:18 PM · Jul 23, 20269.7KViews243560
[AINews] "Laguna S 2.1 Released: Cheaper than Deepseek v4 Flash, Better than V4 Pro" (www.latent.space)

8w deepseek

[AINews] "Laguna S 2.1 Released: Cheaper than Deepseek v4 Flash, Better than V4 Pro" a quiet day lets us highlight a new neolab win. Reignited distillation wars conversation aside, today was more of the same of previous news cycles, which…
SLAI T-Rex: Full-Parameter Post-training of the DeepSeek-V4 Family on Ascend SuperPOD (arxiv.org)

8w moe deepseek

Full-parameter post-training of trillion-parameter-scale MoE models introduces substantial system-level challenges for large-scale distributed training, including severe memory pressure, non-overlapped communication overhead, and inefficie…