Benchmark Almanac

OpenAI Deployment Simulation

GPT-6 Sol 0.08% leads since Sep 2026 Only in system cards

Replays representative production or internal Codex traffic on a candidate model to forecast misaligned or disallowed behavior before launch.

Made by OpenAI Deception and alignment Tasks kept private Last in a system card Oct 2026 Last model added Oct 2026
Scored
Graded by a model A language model scores each response, usually against a rubric.
Answers
Agent task Many steps in a terminal, browser, desktop, app, game or robot.
5
Models scored in lab cards
7
System cards report it, from 1 lab

GPT-6 Sol leads with 0.08%

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

  1. 0.08 GPT-6 Sol
  2. 0.06 GPT-6 Astra
  3. 0.06 GPT-6.1 Sol
  4. 0.04 GPT-6.1 Terra
  5. 0.01 GPT-5.5
  1. 1 GPT-6 Sol 0.08%
  2. 2 GPT-6 Astra 0.06%
  3. 3 GPT-6.1 Sol 0.06%
  4. 4 GPT-6.1 Terra 0.04%
  5. 5 GPT-5.5 0.01%
#ModelScoreSettingDateSource
1 GPT-6 Sol 0.08% Sep 2026 Model card
2 GPT-6 Astra 0.06% Sep 2026 Model card
3 GPT-6.1 Sol 0.06% Sep 2026 Model card
4 GPT-6.1 Terra 0.04% Oct 2026 Model card
5 GPT-5.5 0.01% Apr 2026 Model card

3 models have held number one since 2023

Lab claims count from the day the card came out, independent runs from the model’s release.

OpenAI
Lab claims 3 leaders
GPT-6 Sol 0.08%

Scores over time

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 Best so far
0% 20% 40% 60% 80% 100% 2026 2027 Near the ceiling (85%) GPT-6 Sol GPT-5.5 0% 20% 40% 60% 80% 100% 2026 2027 Near the ceiling (85%) GPT-6 Sol

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.

7 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.

DateDocumentScoreAs the document writes itSection
26 Oct 2026 GPT-6.1 Terra system card addendum 0.045 Deployment Simulation of Internal Codex Traffic, severity 3+ flags, 22 of 48,690 matched tasks (Figure 12, 7.2.2) Alignment
29 Sep 2026 GPT-6.1 Sol system card addendum 0.056 Deployment Simulation of Internal Codex Traffic, severity 3+ flags, 28 of 49,650 tasks (7.6) Alignment
22 Sep 2026 GPT-6 Sol system card 0.083 Deployment Simulation of Internal Codex Traffic, severity 3+ flags, 42 of 50,319 tasks (A.6.5) Alignment
3 Sep 2026 GPT-6 Astra system card 0.063 Deployment Simulation of Internal Codex Traffic, severity 3+ flags, 34 of 54,218 tasks (8.6) Alignment
9 Jul 2026 GPT-5.6 Sol system card Deployment Simulation of Internal Traffic, severity 3 misalignment proportions by type, no overall score (Figure 7, 7.2) Alignment
25 Jun 2026 GPT-5.6 Sol system card Deployment Simulation of Internal Traffic, severity 3 misalignment proportions by type, no overall score (Figure 7, 7.2) Alignment
23 Apr 2026 GPT-5.5 system card 0.01 Resampling internal agentic coding traffic, severity 3 rate, text (category chart in Figure 5, 7.2) Alignment

More in Safety and alignment

All 444 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%
  • AgentDojo In use Agents do tool-use tasks while tools and data carry prompt injections. Attack success rate counts hijacked runs. Grok 4 Fast 0%
  • MASK In use Measures whether a model contradicts its own elicited beliefs under pressure to lie. Honesty separated from accuracy. Muse Spark Open 98.1%
  • 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. GPT-6 Astra 8.50%
  • Agentic misalignment suite (Anthropic) In use Agents in simulated company settings face threats or goal conflicts that invite blackmail, leaking or sabotage. Score is the misaligned action rate. Claude Sonnet 4.5 0%