Claim · 2026-05 · revision · Stated in source · Supported

The Gemini 3 Pro model card states that the model was evaluated as set out in Google DeepMind's Frontier Safety Framework of September 2025 and did not reach any critical capability level, and that in the CBRN domain the uplift level 1 critical capability level was not reached.

One recorded claim. The claim panel gives its qualification, the source passages with their place in each source, and how it was checked.

What it connects

Gemini 3 Pro Model Card → states evaluation result → The card states that Gemini 3 Pro was evaluated under the September 2025 Frontier Safety Framework and reached no critical capability level, with the CBRN uplift level 1 critical capability level recorded as not reached.

Where it appears

No collection page holds this claim. It is listed with its records in the explorer.

Claim and evidence

Gemini 3 Pro Model Cardstates evaluation result ↓The card states that Gemini 3 Pro was evaluated under the September 2025 Frontier Safety Framework and reached no critical capability level, with the CBRN uplift level 1 critical capability level recorded as not reached.

Machine-checked, not reviewed by an editorStated in sourceSupported

Supporting evidence: primary · 1 source

Subject version: The expression stating 'Last Updated: May 2026' · Publication date unknown · Effective date not established · Revised 2026-05
Date: 2026-05 · revision
Evaluated system: Gemini 3 Pro, with the card stating that evaluations using Deep Think mode gave results consistent with the original assessment
What the report measured: Whether the model reaches the critical capability levels of the Frontier Safety Framework, CBRN among the domains
By what kind of method: Frontier safety evaluations as set out in the framework, reported in the card as domain-level key results
The result, in the report’s words: No critical capability level reached; for CBRN the card states the model provides accurate and occasionally actionable information but generally fails to offer novel or sufficiently complete and detailed instructions to significantly enhance the capabilities of low to medium resourced threat actors
The caveats the report gives: The card is a summary that refers to a separate Frontier Safety Framework report for details, and states that model cards may be updated from time to time.

Scope and limitsA developer's determination against its own framework's capability levels, for the model and the version of the framework the card names. It is not an independent finding, and the card refers to a separate Frontier Safety Framework report that was not retrieved (the address tried returned HTTP 404). The card's CBRN cell is quoted because it is the stated result; it describes what the model does not provide and contains no technical content.

The evidence 02

primary sourceSupports the claim

Gemini 3 Pro Model Card

Google DeepMind · Publication date unknown

Cite the original source

Cite the original document, not Selekoa. Check these details against it.

Source excerpt

“We evaluated Gemini 3 Pro as outlined in our latest Frontier Safety Framework (September-2025), and found that it did not reach any critical capability levels as outlined in the table below”

Location in source PDF p. 9, 'Frontier Safety' section

How it was read Retrieved directly from the publisher

Machine-checked · 2026-10-30

Source details and check notes

Authors: Google DeepMind
Version: Retrieved for the candidate pack on 2026-10-28 (SHA-256 9a1734bc3aa310691772d1756990b01383dc7732d4f9cb0a2888e2e06a7f84d8). PDF stating 'Model Release: November 2025, Last Updated: May 2026', as served on 2026-10-28. The deepmind.google model pages render their links in JavaScript, so the card was retrieved from the storage.googleapis.com asset host. Retrieved again at admission on 2026-10-30 (SHA-256 9a1734bc3aa310691772d1756990b01383dc7732d4f9cb0a2888e2e06a7f84d8, the same bytes).
Canonical address: Open the canonical source ↗
Accessed: 2026-10-28
Language: English (en)
Attribution: Google DeepMind, “Gemini 3 Pro Model Card”.
Reuse: No open redistribution licence asserted. Bibliographic metadata, editorial paraphrases and short attributed quotations (at most 50 words each). Recheck terms before wider redistribution.
Check notes: Candidate pack passage, AI and biosecurity (research/ai-bio). The quotation (31 words) matched the text retrieved for the candidate pack on 2026-10-28 and again at admission on 2026-10-30, in a fresh retrieval whose SHA-256 equals the pack’s. Matching ignored differences of whitespace, case, quotation marks, dashes and Unicode compatibility forms, so the source’s typography can differ. A cross-family machine check (2026-10-30, item 44) re-read the record’s sources and found full support. No human editorial review.

primary sourceSupports the claim

Gemini 3 Pro Model Card

Google DeepMind · Publication date unknown

Cite the original source

Cite the original document, not Selekoa. Check these details against it.

Source excerpt

“Gemini 3 Pro provides accurate and occasionally actionable information but generally fails to offer novel or sufficiently complete and detailed instructions to significantly enhance the capabilities of low to medium resourced threat actors”

Location in source PDF p. 10, frontier safety table, CBRN row

How it was read Retrieved directly from the publisher

Machine-checked · 2026-10-30

Source details and check notes

Authors: Google DeepMind
Version: Retrieved for the candidate pack on 2026-10-28 (SHA-256 9a1734bc3aa310691772d1756990b01383dc7732d4f9cb0a2888e2e06a7f84d8). PDF stating 'Model Release: November 2025, Last Updated: May 2026', as served on 2026-10-28. The deepmind.google model pages render their links in JavaScript, so the card was retrieved from the storage.googleapis.com asset host. Retrieved again at admission on 2026-10-30 (SHA-256 9a1734bc3aa310691772d1756990b01383dc7732d4f9cb0a2888e2e06a7f84d8, the same bytes).
Canonical address: Open the canonical source ↗
Accessed: 2026-10-28
Language: English (en)
Attribution: Google DeepMind, “Gemini 3 Pro Model Card”.
Reuse: No open redistribution licence asserted. Bibliographic metadata, editorial paraphrases and short attributed quotations (at most 50 words each). Recheck terms before wider redistribution.
Check notes: Candidate pack passage, AI and biosecurity (research/ai-bio). The quotation (33 words) matched the text retrieved for the candidate pack on 2026-10-28 and again at admission on 2026-10-30, in a fresh retrieval whose SHA-256 equals the pack’s. Matching ignored differences of whitespace, case, quotation marks, dashes and Unicode compatibility forms, so the source’s typography can differ. The row records the critical capability level as 'Uplift Level 1' and 'CCL not reached'. A cross-family machine check (2026-10-30, item 44) re-read the record’s sources and found full support. No human editorial review.

Status and meaning

Editorial status. Not reviewed by an editor. An exploratory record that no editor has accepted. This status does not say whether a person has read any linked passage. Each evidence record above shows how it was checked.

Relation. A report states the result of an evaluation it ran or commissioned, in its own words, with what was measured, by what kind of method, and the caveats the report itself gives (schema 2.22). The record is the report’s statement, not a verified finding: it says nothing about another system, a later version or real-world risk, and a null or limiting result is not evidence that no effect exists.

Date. As stated in the card expression last updated May 2026; the card gives no dates for the evaluations themselves.

Record ID ab-ev-gemini3-cbrn-ccl-result