Humanity's Last Exam On Humanity's Last Exam, a recently released evaluation that tests Al across a broad range of subjects on expert-level questions, the model powering deep research scores a new high at 26.6% accuracy. This test consists of over 3,000 multiple choice and short answer questions across more than 100 subjects from linguistics to rocket science, classics to ecology. Compared to OpenAI 01, the largest gains appeared in chemistry, humanities and social sciences, and mathematics. The model powering deep research showcased a human-like approach by effectively seeking out specialized information when necessary. Model GPT-40 Accuracy (%) 3.3 Grok-2 3.8 Claude 3.5 Sonnet 4.3 Gemini Thinking 6.2 OpenAl 01 9.1 DeepSeek-R1* OpenAl 03-mini (medium)* OpenAl 03-mini (high)* OpenAl deep research** *Model is not multi-modal, evaluated on text-only subset. **with browsing + python tools 9.4 10.5 13.0 26.6

Share Pin Download More

comparison deepseek deep research models

Origin Entry: DeepSeek

File type: png

Resolution: (675px x 685px)

File size: 130 KB

NVIDIA. YOU WERE SUPPOSED TO LEAD THE AI REVOLUTION, NOT DESTROY OUR STOCK PRICE! deepseek I HAVE REMOVED PAYWALLS FROM MY NEW EMPIRE

$11/ الو 111 111 111 AUT 111 \17 1/ الاد ALT 11/ DeepSeek 317 MEX 1/ OpenAl$

Erik Uden @ErikUden@mastodon.de OpenAl in front of the Supreme Court: sorry,,, we needz to use all copyrighted material on earth, otherwise our systems simply don't work uwu OpenAl when another Al company (allegedly) does the same: THEY STOLE OUR WORK? OUR HARD WORK?? STOLEN! CAN YOU BELIEVE IT? 29 janv. 2025, 08:39 - Megalodon 389 625

View All 53 Images

Moar:

0

DeepSeek - It scored rather well compared to openai models

PROTIP: Press the ← and → keys to navigate the gallery, 'g' to view the gallery, or 'r' to view a random image.

Humanity's Last Exam On Humanity's Last Exam, a recently released evaluation that tests Al across a broad range of subjects on expert-level questions, the model powering deep research scores a new high at 26.6% accuracy. This test consists of over 3,000 multiple choice and short answer questions across more than 100 subjects from linguistics to rocket science, classics to ecology. Compared to OpenAI 01, the largest gains appeared in chemistry, humanities and social sciences, and mathematics. The model powering deep research showcased a human-like approach by effectively seeking out specialized information when necessary. Model GPT-40 Accuracy (%) 3.3 Grok-2 3.8 Claude 3.5 Sonnet 4.3 Gemini Thinking 6.2 OpenAl 01 9.1 DeepSeek-R1* OpenAl 03-mini (medium)* OpenAl 03-mini (high)* OpenAl deep research** *Model is not multi-modal, evaluated on text-only subset. **with browsing + python tools 9.4 10.5 13.0 26.6

View Gallery Random Image

+ Add a Comment

Comments (0)

There are no comments currently available.

Display Comments

Image Details

29 views (4 from today)

Uploaded Feb 05, 2025 at 04:03AM EST

Origin Entry

Tags

Claim Authorship Edit History

About the Uploader

Evilthing

Treasurer & Elocutionist & Media Chauffeur

Textile Embed

!https://i.kym-cdn.com/photos/images/newsfeed/003/002/160/f7f.png!

Today's Top Image Galleries

Two edits of JD Vance

J.D. Vance Face Edits

Spooky Pokemon trainer

Hex Maniac

Merrivius Elf and Human Comics example panels.

Merrivius Elf and Human Comics

cartoon illustration of Luigi and Princess Peach playing tennis while Mario and a female Bowser walk by

Bowsette