Benchmark Almanac

Biology Model Refusal Evaluation

Only in system cards

Models answer severe, dual-use and benign biology prompts. Scored as safe rate on harmful prompts and not over-refusing benign ones.

In-house OpenAI safeguard evaluation. Cards report rows with no overall number.

Made by
OpenAI
Released
June 2026
Kind
Refusing harmful requests
Level
Expert
Scored
Graded by a model
Answers
Open-ended
Tasks
Kept private
Measures
Propensity
Language
English
In and out
Text → Text
Interaction
Single turn
Refresh
One fixed set
Last in a system card
November 2026
Last model added
November 2026
1
Model scored in lab cards
19
System cards report it, from 1 lab

GPT-6 Cyber leads with 99.9%

Best score per model as labs report them in their own cards, drawn striped. Labs pick their own settings, so these compare less cleanly than independent runs.

  1. 1 GPT-6 Cyber 99.9%

Source: Labs’ model cards

#ModelScoreSettingDateSource
1 GPT-6 Cyber 99.9% Nov 2026 Lab model card

OpenAI’s own number for GPT-6 Cyber passed 85% in November 2026

That was 5 months after its release. By the same age, 25% of the benchmarks from 2026 had passed 85%. All the scores we have, one dot each, so a model tested at several settings appears more than once.

Reported by the lab in its model card
0% 20% 40% 60% 80% 100% 2027 Near the ceiling (85%) GPT-6 Cyber 0% 20% 40% 60% 80% 100% 2027 Near the ceiling (85%) GPT-6 Cyber

Source: Labs’ model cards

Labs report it only in system cards

We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark. Some system cards do, below.

19 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. OpenAI accounts for 11 of the 11 releases that named it since 2024.

DateDocumentScoreAs the document writes itSection
Nov 17, 2026 GPT-6.2 Sol system card Biology Model Refusal Evaluation, severe, dual-use and benign rows, no overall score (Table 6) Safety
Nov 17, 2026 GPT-6.2 Astra system card Biology Model Refusal Evaluation, severe, dual-use and benign rows, no overall score (Table 6) Safety
Nov 5, 2026 GPT-6 Cyber system card 99.9 Biology refusal evaluations, severe-risk requests, safe responses 99.87% (dual-use 98.27%, benign no unnecessary refusal 97.22%) Bio and chem risk
Oct 29, 2026 GPT-6 Luna system card addendum Biology Model Refusal Evaluation, severe and dual use rows, no overall score (Table 10) Safety
Oct 26, 2026 GPT-6.1 Terra system card addendum Biology Model Refusal Evaluation, severe 0.990, dual use 0.960, benign not over-refuse 0.982, no overall score (Table 31) Safety
Oct 7, 2026 GPT-6 Sol system card addendum Biology Model Refusal Evaluation, severe and dual use rows, no overall score (Table 10) Safety
Oct 7, 2026 GPT-6 Luna system card addendum Biology Model Refusal Evaluation, severe and dual use rows, no overall score (Table 10) Safety
Sep 29, 2026 GPT-6.1 Sol system card addendum Biology Model Refusal Evaluation, severe, dual use and benign rows, no overall score (Table 11) Safety
Sep 22, 2026 GPT-6 Sol system card Biology Model Refusal Evaluation, severe, dual use and benign rows, no overall score (Table 35) Safety
Sep 22, 2026 GPT-6 Luna system card Biology Model Refusal Evaluation, severe, dual use and benign rows, no overall score (Table 35) Safety
Sep 3, 2026 GPT-6 Astra system card Biology Model Refusal Evaluation, severe, dual use and benign rows, no overall score (Table 19) Safety
Aug 6, 2026 GPT-5.6 Sol system card addendum Biology Model Refusal Evaluation, severe and dual use rows (0.954, 0.945), no overall score (7.2.1.1) Safety
Aug 6, 2026 GPT-5.6 Luna system card addendum Biology Model Refusal Evaluation, severe and dual use rows (0.937, 0.928), no overall score (7.2.1.1) Safety
Jul 9, 2026 GPT-5.6 Terra system card Biology Model Refusal Evaluation, severe, dual use and benign rows (0.950, 0.911, 0.978), no overall score (9.4.2.1) Safety
Jul 9, 2026 GPT-5.6 Sol system card Biology Model Refusal Evaluation, severe, dual use and benign rows (0.943, 0.911, 0.989), no overall score (9.4.2.1) Safety
Jul 9, 2026 GPT-5.6 Luna system card Biology Model Refusal Evaluation, severe, dual use and benign rows (0.946, 0.926, 0.989), no overall score (9.4.2.1) Safety
Jun 25, 2026 GPT-5.6 Terra system card Biology Model Refusal Evaluation, severe, dual use and benign rows (0.950, 0.911, 0.978), no overall score (9.3.2.1) Safety
Jun 25, 2026 GPT-5.6 Sol system card Biology Model Refusal Evaluation, severe, dual use and benign rows (0.943, 0.911, 0.989), no overall score (9.3.2.1) Safety
Jun 25, 2026 GPT-5.6 Luna system card Biology Model Refusal Evaluation, severe, dual use and benign rows (0.946, 0.926, 0.989), no overall score (9.3.2.1) Safety

More in Safety and alignment

All 636 in Safety and alignment
  • FORTRESS In use Paired adversarial and benign national-security misuse prompts. Rubric-graded risk score on adversarial and over-refusal score on benign prompts. Claude Opus 5.5 5.78%
  • StrongREJECT In use Forbidden-prompt jailbreak benchmark with a rubric grader that scores how useful a non-refusal actually is. Inkling 98.6%
  • Gray Swan IPI Arena benchmark In use Indirect prompt-injection attacks from public red-teaming competitions replayed on tool, coding and browser agents. Attack success rate at k attempts. Gemini 4 Carbon 0%
  • Covert misbehavior (Thinking Machines) In use Agentic tasks that check whether a model misbehaves and conceals it. Scored by the rate of covert misbehavior. Inkling 2 Mini 8.30%
  • Destructive actions (Thinking Machines) In use Agentic tasks that check whether a model takes destructive actions it was not authorized to take. Scored by the rate of destructive actions. Inkling 2 Small 4.70%
  • Misbehavior (Thinking Machines) In use Agentic tasks that check whether a model fabricates results, ignores instructions or hides what it did when a task is hard. Scored by the rate of misbehavior. Inkling 2 Mini 9.80%

Data audit

All 4 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access The tasks are private.
  • Pass Confidence Checked against a primary source.
Scores 1 pass 4 notes
  • Note Independent runs Only labs’ own numbers, as expected for an in-house eval. Scores from system cards (1 score).
  • Note Models scored Only 1 model we track.
  • Pass Newest score The newest is from November 2026, for GPT-6.2 Astra.
  • Note Frontier models None of its models are on the capabilities index.
  • Note Headroom Saturated. The best is 99.9%, past 85% since November 2026.
Leaderboard import 1 note
  • Note Importer No public leaderboard to read.
Consistency 1 pass 1 note
  • Note Lab claims No model has both a lab number and an independent run.
  • Pass Known issues No published audit found problems.

How we check