Skip to content

Instantly share code, notes, and snippets.

@wilmoore

wilmoore/readme.md Secret

Last active January 30, 2025 21:15
Show Gist options

  • Select an option

    Learn more about clone URLs
  • Save wilmoore/9f57e1f1d3fb242ec7506e1b150f6ce9 to your computer and use it in GitHub Desktop.

Select an option

Learn more about clone URLs
Save wilmoore/9f57e1f1d3fb242ec7506e1b150f6ce9 to your computer and use it in GitHub Desktop.
Download ZIP
Personal Brand :: Social Media :: LinkedIn :: Post :: πŸ”₯ DeepSeek is killing the game with its performance! πŸ”₯
Raw
readme.md

Personal Brand :: Social Media :: LinkedIn :: Post :: πŸ”₯ DeepSeek is killing the game with its performance! πŸ”₯

βͺΌ Made with πŸ’œ by Polyglot.

assistant
reference
related

πŸ”₯ DeepSeek is killing the game with its performance! πŸ”₯

πŸ”₯ DeepSeek is killing the game with its performance! πŸ”₯

The DeepSeek family of models is setting new benchmarks and leaving the competition in the dust...

Let’s dive into the raw numbers:

πŸ’₯ AIME 2024 (Pass@1)
DeepSeek-R1 dominates with an impressive 79.8%, showcasing its exceptional problem-solving prowess.

πŸ’₯ Codeforces (Percentile)
Competitive programming? DeepSeek-R1-32B crushes it at 96.3%, raising the bar for others to follow.

πŸ’₯ GPQA Diamond (Pass@1)
Tackling intricate QA challenges, DeepSeek-R1 shines at 71.5%.

πŸ’₯ MATH-500 (Pass@1)
DeepSeek-R1 flexes its mathematical reasoning muscles, achieving a stellar 97.3%.

πŸ’₯ MMLU (Pass@1)
When it comes to mastering diverse language tasks, DeepSeek-R1 delivers with a solid 90.8%.

πŸ’₯ SWE-bench Verified (Resolved)
DeepSeek-V3 leads the way, excelling in software engineering tasks at 49.2%, redefining what’s possible in technical problem-solving.


OpenAI’s models are solid performers but are clearly trailing the momentum of DeepSeek.


ȚechΕ›avvy CEO

image

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment