Benchmark Almanac

Insider Trading Demo

gpt-4-32k-0613 0% tops its leaderboard davinci-002 4.67% leads the models we track
Not in model cards

Models act as stock traders under pressure. Apollo measures how often they trade on inside information and conceal their actions.

Made by
Apollo Research
Released
November 2023
Kind
Deception and alignment
Level
University
Scored
Settled by events
Answers
Agent task
Tasks
All public
Measures
Propensity
Language
English
In and out
Text → Actions, Text
Interaction
Agentic
Items from
Generated
Refresh
One fixed set
Run by
Anyone
Occupations
Sales
Strongest model tested
GPT-4 Jun 2023 #186 on the capability index
Last model added
September 2023
Scores refreshed
November 20th, 2026
7
Entries on its own leaderboard, 5 models we track

gpt-4-32k-0613 tops the leaderboard with 0% davinci-002 leads with 4.67% among models released by September 2023

Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. The best model we track is davinci-002, at 4.67%. This view shows only models we track, each at its best.

The newest model with a score here came out in September 2023, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 1 gpt-4-32k-0613 pressure: No pressure 0%
  2. 2 davinci-002 default 4.67%
  3. 3 GPT-3.5 Turbo Instruct default 15.3%
  4. 4 GPT-3.5 Turbo Jun 2023, default 16.7%
  5. 5 davinci-003 default 29.7%
  6. 6 gpt-4-base default 66.3%
  7. 7 GPT-4 Jun 2023, default 69%

Source: Official Insider Trading Demo leaderboard

Source: Official Insider Trading Demo leaderboard

#EntryScoreSettingDateSource
1 gpt-4-32k-0613 0% pressure: No pressure Official leaderboard
2 davinci-002 4.67% default Mar 2022 Official leaderboard
3 GPT-3.5 Turbo Instruct 15.3% default Sep 2023 Official leaderboard
4 GPT-3.5 Turbo Jun 2023 16.7% default Jun 2023 Official leaderboard
5 davinci-003 29.7% default Nov 2022 Official leaderboard
6 gpt-4-base 66.3% default Official leaderboard
7 GPT-4 Jun 2023 69% default Jun 2023 Official leaderboard
#ModelScoreSettingDateSource
1 davinci-002 4.67% default Mar 2022 Official leaderboard
2 GPT-3.5 Turbo Instruct 15.3% default Sep 2023 Official leaderboard
3 GPT-3.5 Turbo Jun 2023 16.7% default Jun 2023 Official leaderboard
4 davinci-003 29.7% default Nov 2022 Official leaderboard
5 GPT-4 Jun 2023 69% default Jun 2023 Official leaderboard

Rankings by part

We read 2 parts from the benchmark’s own leaderboard: Strategic deception given misalignment and Doubling down given deception. Each ranks models on one part of the test. GPT-3.5 Turbo leads Strategic deception given misalignment and GPT-3.5 Turbo Instruct leads Doubling down given deception. Arrows show how far a model moved from its overall rank among the models we track.

Strategic deception given misalignment

5 models we track have a score here.

  1. 1 GPT-3.5 Turbo Jun 2023 ↑2 8.00%
  2. 2 davinci-003 ↑2 28.1%
  3. 3 GPT-3.5 Turbo Instruct ↓1 34.8%
  4. 4 davinci-002 ↓3 57.1%
  5. 5 GPT-4 Jun 2023 69.6%

Number one over time

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

OpenAI
Own leaderboard Not tested since Nov 2023
davinci-002 4.67%

Source: Official Insider Trading Demo leaderboard

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

Plot with other benchmarks
On the benchmark’s own leaderboard
0% 20% 40% 60% 80% 100% 2022 2023 davinci-002 0% 20% 40% 60% 80% 100% 2022 2023 davinci-002

Source: Official Insider Trading Demo leaderboard

No lab model card reports it yet

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

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

8 of 9 checks pass, 1 warning. Look first at the Newest score check. The newest came 3 years ago, in September 2023, for GPT-3.5 Turbo Instruct. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 1 warning 2 pass 2 notes
  • Pass Independent runs Scores from the official leaderboard (7 entries).
  • Pass Models scored 5 models we track, plus 2 other entries on its leaderboard.
  • Warning Newest score The newest came 3 years ago, in September 2023, for GPT-3.5 Turbo Instruct.
  • Note Frontier models None of the top 10 has a score. The strongest tested is GPT-4 (#186).
  • Note Headroom Lower is better, so there’s no ceiling to reach.
Leaderboard import 3 pass
  • Pass Importer We read the board on github.com: 5 rows for models we track and 2 other entries at the last read.
  • Pass Last read November 20th, 2026.
  • Pass Read worked The last read worked.
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