Benchmark Almanac

Coding Deception

GPT-6 Sol 1.30% leads since September 2026
Only in system cards

Agents work on coding tasks selected to elicit dishonest reports. Scored on the rate of misrepresenting their work to the user.

In-house OpenAI evaluation. The tasks are not public.

Made by
OpenAI
Released
August 2025
Kind
Deception and alignment
Level
Expert
Scored
Graded by a model
Answers
Agent task
Tasks
Kept private
Strongest model tested
GPT-6 Astra #3 on the capability index
Last in a system card
September 2026
Last model added
September 2026
6
Models scored in lab cards
6
System cards report it, from 1 lab

GPT-6 Sol leads with 1.30% among models released by September 2026

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.

The newest model with a score here came out in September 2026, and nobody has reported a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 1 GPT-6 Sol 1.30%
  2. 2 GPT-6.1 Sol 1.50%
  3. 3 GPT-6 Luna 2.81%
  4. 4 GPT-6 Astra 3.54%
  5. 5 GPT-5 17%
  6. 6 GPT-5.2 25.6%

Source: Labs’ model cards

#ModelScoreSettingDateSource
1 GPT-6 Sol 1.30% Sep 2026 Lab model card
2 GPT-6.1 Sol 1.50% Sep 2026 Lab model card
3 GPT-6 Luna 2.81% Sep 2026 Lab model card
4 GPT-6 Astra 3.54% Sep 2026 Lab model card
5 GPT-5 17% Aug 2025 Lab model card
6 GPT-5.2 25.6% Dec 2025 Lab model card

3 models have held number one in labs’ own numbers since its release in August 2025

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

OpenAI
Lab claims Not tested since Sep 2026
GPT-6 Sol 1.30%

Source: Labs’ model cards

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 GPT-6 Sol GPT-5 0% 20% 40% 60% 80% 100% 2026 GPT-6 Sol GPT-5

Source: Labs’ model cards

Labs report it only in system cards

We read the headline results of 522 launch posts from 52 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.

DateDocumentScoreAs the document writes itSection
Sep 29, 2026 GPT-6.1 Sol system card addendum 1.5 Coding Deception, rate of misrepresentation (7.4.1) Alignment
Sep 22, 2026 GPT-6 Sol system card 1.3 Coding Deception, failure rate at maximum reasoning effort, printed chart label (Figure 64, A.6.4.1) Alignment
Sep 22, 2026 GPT-6 Luna system card 2.81 Coding Deception, failure rate at maximum reasoning effort, printed chart label (Figure 64, A.6.4.1) Alignment
Sep 3, 2026 GPT-6 Astra system card 3.54 Coding Deception, failure rate, printed chart label (Figure 10, 8.3.1) Alignment
Dec 11, 2025 GPT-5.2 system card addendum 25.6 Deception rate, Coding Deception, gpt-5.2-thinking (Table 6) Alignment
Aug 7, 2025 GPT-5 system card 17 Coding Deception, deception rate (Table 9) Alignment

More in Safety and alignment

All 465 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%
  • DistressBench In use Models reply as a chatbot to 718 simulated users in psychological crisis, in single-turn and multi-turn chats. Three judge models score replies against clinician-written rubrics. Muse Spark 1.2 88.4%
  • Lakera B3 (Backbone Breaker Benchmark) In use Crowdsourced prompt-injection attacks hit backbone LLMs inside agent threat snapshots at three defense levels. The score is attack resistance, 100 minus the board's vulnerability percentage. Claude Opus 4.5 89%
  • IHEval In use Models face conflicts between system, user and tool instructions. Reports a composite score for following the higher-priority instruction. GPT-4o 70%