Vibeleaderboard
← All Intel
Intel / post

Congrats to @deepseek_ai for their release of DeepSeekv4 Pro 0813 1.5T🔥 It…

Source
x.com
Date
SemiAnalysis@SemiAnalysis_

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 parameters! There are lots of brilliant people working on Nemotron, but fundamentally, committee-based model frontier development does not work.👇️ 1/3🧵

View on X

Context

SemiAnalysis's comparison chart places DeepSeek's V4 Flash 0731 and V4 Pro 0813 ahead of NVIDIA's Nemotron 3 Ultra NVFP4 on Terminal-Bench 2.1, a test of terminal-based task execution. The displayed scores are 82.7%, 87.9% and 53.9%, respectively.

These are vendor-reported results: the chart cites NVIDIA's Nemotron 3 Ultra model card and DeepSeek's official table. DeepSeek's model card lists the Pro score and specifies that its code- tests used DeepSeek's own agent framework at max reasoning effort. The comparison concerns those tested systems; it does not isolate model architecture or establish the post's proposed organizational explanation for the gap.

Terms in this piece · Glossary
  • benchmark — A standard public test set for comparing AI models — the shared scoreboards behind every "model X beats model Y" claim.
  • AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
Read the source x.com
More from SemiAnalysis
Recommended reads
Comments

Checking sign-in…

Loading comments…