WebGPT 175B
A GPT-3 175B-derived research system trained with human feedback to search and navigate a text browser, collect references, and compose cited long-form answers.
0 comments · 0 votes
Sign in to join the discussion →
No comments yet. Start the discussion.
Why this model scores 33.0
Broad web-assisted question answering and multi-step browser use raise capability and autonomy over base GPT-3, while research-only access, a restricted GET-style environment, citation visibility, and no public release keep deployment and residual control difficulty low.
News tied to WebGPT 175B
The model score of 33.0 rates this model's risk profile. The overall Doom Index of 66.8 measures the complete temporally weighted evidence record. These values answer different questions.
Each article's current Doom Index contribution is divided equally among the exact models named on that article. This prevents multi-model evidence from being claimed in full on several model pages. Model risk scores use a bounded temporal offset around their technical profile, but never feed back into the overall index.
OpenAI WebGPT browses the web and outperforms human demonstrators
WebGPT autonomously issued searches, followed links, scrolled pages, gathered citations, and produced long-form answers. Its best 175B system was preferred over human demonstrations 56 percent of the time, and later motivated deployed ChatGPT browsing plugins.
- Full item contribution
- +0.03
- WebGPT 175B equal share
- +0.03
Model score history
-
R3
Doom Score 33.0
Source-backed model availability audit using the model's existing primary or authoritative catalogue evidence. Exact-version evidence chronology replayed under temporal-monthly-pressure-v4.
12 Aug 2026 -
R2
Doom Score 33.0
Exact-version evidence chronology replayed after run doombench-hourly-news-20260812-142416 under temporal-monthly-pressure-v4.
12 Aug 2026 -
R1
Doom Score 35.1
New exact research-model profile distinct from GPT-3 175B, supported by OpenAI's release and paper; dimensions reflect verified tool use and restricted research-only access.
12 Aug 2026