CBRN expert red teaming
Only in system cards
Experts try to extract uplift beyond a web baseline across staged CBRN attack scenarios.
In-house Google DeepMind evaluation. Results are usually qualitative or rubric scores per scenario.
Made by Google DeepMind Released Nov 2025 Hazardous knowledge Tasks kept private Last in a system card Oct 2026
Last model added Oct 2026
- Level
- Expert Specialists with years of training, often over hours of work
- Scored
- Graded by people Human experts grade or compare the outputs.
- Answers
- Open-ended Free-form text, documents, designs or media judged as a whole.
6
System cards report it, from 1 lab
No scores here yet
No lab, leaderboard or Epoch AI run has reported a score for CBRN expert red teaming yet.
Labs report it only in system cards
We read the headline results of 479 launch posts from 51 labs, and none of them list this benchmark. Some system cards do, below.
6 system cards report it, from 1 lab
Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.
| Date | Document | Score | As the document writes it | Section |
|---|---|---|---|---|
| 13 Oct 2026 |  Gemini 4 Argon system card | Biological expert red teaming, mean 0-100 rubric score per scenario, no overall: BIO-1 91.2, BIO-2 88.5, BIO-3 73.1 vs web baselines 69.6, 62.3, 75.3 (Table 3.3, 3.2.1.3). Chemical, radiological and nuclear red teaming reported qualitatively (3.2.1.4) | Bio and chem risk | |
| 13 Oct 2026 | Gemini 4 Argon model card | CBRN expert red teaming: measurable uplift over web baselines in two biological scenarios, biological CCL cannot be ruled out, no score (Frontier Safety table, p.11) | Bio and chem risk | |
| 13 Aug 2026 | Gemini 3.7 Flash model card | CBRN expert red teaming: modest uplift over web baselines, alert threshold reached, CCL not reached, no score (Frontier Safety table, p.8) | Bio and chem risk | |
| 13 Aug 2026 | Gemini 3.7 Flash system card | Expert red teaming, aggregated 0-100 rubric scores vs web baseline: BIO-1 77.6 (web 69.6), BIO-2 79.1 (web 67.0), chemical scenarios 71.1 (web 47.0) and 68.9 (web 41.0). No overall score (CBRN, Results, p.10 and 13) | Bio and chem risk | |
| 19 Feb 2026 | Gemini 3.1 Pro model card | CBRN (Deep Think mode): accurate and actionable information but not novel or complete enough for critical stages, CCL not reached, no score (FSF results table, p.8) | Bio and chem risk | |
| 18 Nov 2025 | Gemini 3 Pro system card | Red-Teaming: minimal uplift over web baseline across all four domains, no score (CBRN, p.7 to 8) | Bio and chem risk |
More in Biology and chemistry
- SUPERChem In use Models solve 500 expert-written chemistry reasoning problems, offered in multimodal and text-only forms. Expert solution paths also allow a reasoning path score.
Tencent Hy4 preview 66.4% - GeneBench Pro In use Agents analyze messy simulated genomics and biology datasets to estimate a decision-relevant quantity across 129 problems. Scored automatically against known answers. GPT-6 Astra 37.1%
- LifeSciBench In use Models complete 750 expert-written life science research tasks, often with attached figures, tables, or sequence files. Graded against expert rubrics as score and pass rate. GPT-6 Astra 60.3%
- MolecularIQ In use Models answer questions about molecular graphs, such as counting, indexing and constrained generation, and a symbolic checker verifies each answer.
Intern-S2-397B 62.4% - TOMG-Bench In use Models generate molecules from natural language requests in three tasks: molecule editing, molecule optimization and customized generation. Each output is checked for validity and for meeting the requested properties. Intern-S2-397B 66.8%
- BABE Fading Models reason from experimental results in peer-reviewed biology papers to draw conclusions, with causal and cross-scale inference.
Seed2.0 Lite (0428) 57.9%