English(EN) Congrats to @deepseek_ai for their release of DeepSeekv4 Pro 0813 1.5T🔥 It massively beats Nemotron3 Ultra on agentic tasks.

DeepSeek 发布新模型，性能超越 Nemotron3 Ultra

作者 PulseAugur 编辑部 · [2 个来源] · 2026-08-12 18:32

DeepSeek 发布了两款新模型：DeepSeekv4 Pro 0813 1.5T 和 DeepSeekv4 Flash 0731。据称，这两款模型在代理任务上的表现优于 NVIDIA 的 Nemotron3 Ultra。DeepSeekv4 Flash 模型以更少的参数实现了这一目标。分析还表明，Mistral AI 为 Nemotron 所做的预训练工作并不成功，而 Nemotron 的委员会式开发结构正在阻碍其进展。 AI

影响 DeepSeek 的新模型可能会为代理任务设定新的基准，对 NVIDIA 和 Mistral AI 等竞争对手构成压力。

排序理由该集群报道了 DeepSeek 发布新的前沿模型，并声称其性能可与竞争对手模型相媲美。[lever_c_demoted from frontier_release: ic=2 ai=1.0]

在 X — SemiAnalysis 阅读 →

AI 生成摘要 · Google Gemini · 来自 2 个来源。我们如何撰写摘要 →

报道来源 [2]

X — SemiAnalysis TIER_1 English(EN) · SemiAnalysis_ · 2026-08-12 18:32

@deepseek_ai Furthermore, Mistral fumbled Nemotron's pretraining. NVIDIA's single-committee structure means that one member's failure drags everyone down with i

@deepseek_ai Furthermore, Mistral fumbled Nemotron's pretraining. NVIDIA's single-committee structure means that one member's failure drags everyone down with it, and the rest of the committee is upset. 2/3🧵
X — SemiAnalysis TIER_1 English(EN) · SemiAnalysis_ · 2026-08-12 18:32

Congrats to @deepseek_ai for their release of DeepSeekv4 Pro 0813 1.5T🔥 It massively beats Nemotron3 Ultra on agentic tasks.

Congrats to @deepseek_ai for their release of DeepSeekv4 Pro 0813 1.5T🔥 It massively beats Nemotron3 Ultra on agentic tasks. This is in addition to DeepSeekv4 Flash 0731 beating Nemotron3 Ultra massively too, while having 4.2x fewer active parameters and close to 2x fewer total …

报道来源 [2]

@deepseek_ai Furthermore, Mistral fumbled Nemotron's pretraining. NVIDIA's single-committee structure means that one member's failure drags everyone down with i

Congrats to @deepseek_ai for their release of DeepSeekv4 Pro 0813 1.5T🔥 It massively beats Nemotron3 Ultra on agentic tasks.

相关实体

相关话题