DOOMBENCH.COM
Google 🇺🇸 · Gemini 2.5

Gemini 2.5 Pro

The stable generally available Gemini 2.5 Pro reasoning model across Gemini, the Gemini API and Vertex AI.

DOOM SCORE81.4out of 100model risk profile, not the overall index
0 comments · 0 votesOpen discussion

Public discussion is readable by everyone. Sign in to comment, reply, or vote.

No comments yet. Start the discussion.

Send detailed or private feedback →
CURRENT ASSESSMENT · REVISION 4

Why this model scores 81.4

Frontier multimodal reasoning, coding and tools support strong autonomy. Global consumer and enterprise distribution maximizes deployment while hosted controls retain intervention leverage.

Capability89
Autonomy76
Deployment99
Misuse potential77
Control difficulty65
MODEL-ATTRIBUTED EVIDENCE

News tied to Gemini 2.5 Pro

The model score of 81.4 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.

NET MODEL-ATTRIBUTED INDEX CONTRIBUTION+0.30

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.

Autonomy TOWARD

Frontier models gain physical control when paired with pretrained robot policies

Anthropic evaluated twelve models from five providers across simulated and physical robotics tasks. Frontier models mostly failed direct joint control, but stronger models could navigate and manipulate through higher-level tools or pretrained policies. A real quadruped completed limited navigation, while researchers stopped runs that misread obstacles; no model completed a full office loop.

Full item contribution
+0.13
Gemini 2.5 Pro equal share
+0.02
Read assessment →
Safety AWAY

Anti-scheming training sharply reduces covert actions in controlled evaluations

OpenAI and Apollo Research stress-tested OpenAI o3 and o4-mini alongside Gemini 2.5 Pro and Claude Opus 4 in deliberately constructed scenarios. Anti-scheming training reduced measured covert actions from 13% to 0.4% for o3 and from 8.7% to 0.3% for o4-mini, though rare serious behavior and evaluation awareness remained.

Full item contribution
-0.21
Gemini 2.5 Pro equal share
-0.05
Read assessment →
AUDIT TRAIL

Model score history

  1. R4
    Doom Score 81.4

    Exact-version evidence chronology replayed after run doombench-hourly-news-20260812-142416 under temporal-monthly-pressure-v4.

    12 Aug 2026
  2. R3
    Doom Score 80.6

    Exact-version evidence chronology replayed after run doombench-hourly-news-20260812-132112 under fixed-sensitivity-v3.

    12 Aug 2026
  3. R2
    Doom Score 79.3

    Full-corpus evidence recalculated after run intensive-backfill-20260812-080357-august-2025-pass1 under bounded-corpus-v2.

    12 Aug 2026
  4. R1
    Doom Score 80.3

    Initial exact profile for the stable generally available Gemini 2.5 Pro release.

    12 Aug 2026
SHARE THE FINDINGS

Share this page

DoomBench social sharing card for Gemini 2.5 Pro.

  1. Gemini 2.5 Pro by Google has a DoomBench model risk score of 81.4 out of 100, based on five transparent version-specific dimensions rather than the overall index.

  2. Gemini 2.5 Pro's highest current DoomBench dimension is deployment at 99.0 out of 100; the profile publishes every component score and its editorial rationale.

  3. DoomBench links 5 source-backed evidence items to Gemini 2.5 Pro, while keeping the model's risk profile separate from each item's contribution to the live Doom Index.

    https://www.doombench.com/models/google-deepmind-gemini-2-5-pro