GPT-4
A multimodal frontier model that reached human-level performance on several professional and academic benchmarks while retaining material reliability limits.
0 comments · 0 votes
Sign in to join the discussion →
No comments yet. Start the discussion.
Why this model scores 51.4
GPT-4 was a major general-capability step with broad API deployment, but it lacked the sustained agentic autonomy and dangerous-domain performance of later frontier systems.
News tied to GPT-4
The model score of 51.4 rates this model's risk profile. The overall Doom Index of 67.9 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.
Altman forecasts millions of AI coworkers across knowledge work
Altman forecast roughly one million virtual coworkers in every knowledge-work field, rapidly falling intelligence costs, uneven sectoral effects, and a possible breakdown in the balance of power between capital and labor.
- Full item contribution
- +0.09
- GPT-4 equal share
- +0.04
OpenAI details deployed Rule-Based Rewards safeguards
OpenAI published Rule-Based Rewards, a safety-training method used since GPT-4 and in GPT-4o mini that matched human-feedback safety performance while reducing over-refusals and the need for repeated human labeling.
- Full item contribution
- -0.13
- GPT-4 equal share
- -0.06
Andrew Ng shows agent loops can outperform a stronger model's single pass
Andrew Ng synthesized coding evaluations in which GPT-3.5 reached up to 95.1 percent on HumanEval when wrapped in an iterative agent loop, compared with 67.0 percent for zero-shot GPT-4, and identified reflection, tool use, planning, and multi-agent collaboration as transferable capability multipliers.
- Full item contribution
- +0.06
- GPT-4 equal share
- +0.06
ARC Evals finds 2023 agents far below autonomous replication threshold
ARC Evals tested four GPT-4- and Claude-based agents on 12 realistic autonomous replication and adaptation tasks. The agents completed only the easiest tasks and failed end-to-end persistent deployment and controlled phishing attempts, indicating that casual users of those versions were unlikely to create dangerous autonomous agents. OpenAI's GPT-4 system card independently documents ARC's predeployment evaluation role.
- Full item contribution
- -0.08
- GPT-4 equal share
- -0.08
Altman proposes licensing and external validation for frontier AI
In written Senate testimony, Altman proposed licensing or registration above a capability threshold, internal and external testing, published evaluations, disclosure duties, independent validation, and international coordination.
- Full item contribution
- -0.03
- GPT-4 equal share
- -0.03
Rohin Shah warns a short AI pause could reduce safety time
Rohin Shah supports enforceable global slowing, but identifies a countervailing mechanism: hardware, algorithms, and investment could keep advancing during a training pause while safety researchers lose access to intermediate models. A poorly designed pause could therefore reduce capabilities-adjusted safety time without proportionally delaying dangerous systems.
- Full item contribution
- +0.02
- GPT-4 equal share
- +0.02
Auto-GPT launches an open autonomous-agent framework
Significant Gravitas released Auto-GPT as an open experimental application that used GPT-4 to break goals into tasks, call tools and continue working with limited human prompting.
- Full item contribution
- +0.06
- GPT-4 equal share
- +0.06
Eliezer Yudkowsky calls for an enforceable global halt to large AI training runs
Eliezer Yudkowsky argued that a six-month pause would not address loss-of-control risk from smarter-than-human AI. He proposed an indefinite worldwide moratorium on large training runs, lower compute ceilings as algorithms improve, GPU tracking, multinational enforcement, and no government or military exemptions.
- Full item contribution
- +0.03
- GPT-4 equal share
- +0.03
OpenAI releases GPT-4
GPT-4 improved broad reasoning and multimodal capability while documenting important limitations and safety work.
- Full item contribution
- +0.10
- GPT-4 equal share
- +0.10
Model score history
-
R9
Doom Score 51.4
Exact-version evidence chronology replayed after run doombench-hourly-news-20260824-170007 under temporal-monthly-pressure-v4.
24 Aug 2026 -
R8
Doom Score 51.5
Exact-version evidence chronology replayed after run doombench-hourly-news-20260814-051858z under temporal-monthly-pressure-v4.
14 Aug 2026 -
R7
Doom Score 51.4
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 -
R6
Doom Score 51.4
Exact-version evidence chronology replayed after run doombench-hourly-news-20260812-142416 under temporal-monthly-pressure-v4.
12 Aug 2026 -
R5
Doom Score 53.5
Exact-version evidence chronology replayed after run doombench-hourly-news-20260812-132112 under fixed-sensitivity-v3.
12 Aug 2026 -
R4
Doom Score 56.0
Full-corpus evidence recalculated after run intensive-backfill-20260812-040700-july-2024-incremental under bounded-corpus-v2.
12 Aug 2026 -
R3
Doom Score 58.4
Full-corpus evidence recalculated after run intensive-backfill-20260812-010634-march-2023-incremental under bounded-corpus-v2.
12 Aug 2026 -
R2
Doom Score 56.3
Full-corpus evidence recalculated after run intensive-backfill-20260811-191225 under bounded-corpus-v2.
11 Aug 2026 -
R1
Doom Score 51.4
Initial source-backed model assessment
11 Aug 2026