Benchmark Almanac

MakeMePay

GPT-4.5 57% leads since February 2025
Only in system cards

A con artist model tries to talk a GPT-4o mark into paying money in a role-play. Scored as the share of conversations with a payment.

Open source in OpenAI Evals. Cards also report the dollar extraction rate.

Made by
OpenAI
Released
September 2023
Kind
Dangerous capabilities
Level
Everyday
Scored
Settled by events
Answers
Open-ended
Tasks
All public
Strongest model tested
o1 #117 on the capability index
Last in a system card
February 2025
Last model added
February 2025
6
Models scored in lab cards
6
System cards report it, from 1 lab

GPT-4.5 leads with 57% among models released by February 2025

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 February 2025, and nobody has reported a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 57.0 GPT-4.5
  2. 36.0 OpenAI deep research
  3. 27.0 o1
  4. 11.6 o1-preview
  5. 1.00 o3-mini
  6. 0.70 o1-mini
  1. 1 GPT-4.5 57%
  2. 2 OpenAI deep research 36%
  3. 3 o1 27%
  4. 4 o1-preview 11.6%
  5. 5 o3-mini 1.00%
  6. 6 o1-mini 0.70%

Source: Labs’ model cards

#ModelScoreSettingDateSource
1 GPT-4.5 57% Feb 2025 Lab model card
2 OpenAI deep research 36% Feb 2025 Lab model card
3 o1 27% Dec 2024 Lab model card
4 o1-preview 11.6% Sep 2024 Lab model card
5 o3-mini 1.00% Jan 2025 Lab model card
6 o1-mini 0.70% Sep 2024 Lab model card

4 models have held number one in labs’ own numbers since its release in September 2023

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

OpenAI
Lab claims Not tested since Feb 2025
GPT-4.5 57%

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% 2025 Near the ceiling (85%) GPT-4.5 OpenAI deep research o1 o1-preview 0% 20% 40% 60% 80% 100% 2025 Near the ceiling (85%) GPT-4.5 OpenAI deep research

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
Feb 27, 2025 GPT-4.5 system card 57 MakeMePay vs GPT-4o, % of times con-artist model received payment, post-mitigation, dollar extraction rate 8% (4.5.1) Persuasion
Feb 25, 2025 OpenAI deep research system card 36 MakeMePay vs GPT-4o, % of times con-artist model received payment, post-mitigation without browsing (dollar extraction 21%) (3.4.6) Persuasion
Jan 31, 2025 o3-mini system card 1 MakeMePay vs GPT-4o, % of times con-artist model received payment, o3-mini (Post), dollar extraction 0%, printed chart label (5.6.3) Persuasion
Dec 5, 2024 o1 system card 27 MakeMePay vs GPT-4o, % of times con-artist model received payment, o1 (post-mitigation) (dollar extraction 4%) (5.7.3) Persuasion
Sep 12, 2024 o1-preview system card 11.6 MakeMePay, % of times con-artist model received a donation from GPT-4o, o1-preview (post-mitigation) (dollar extraction 3.2%) (4.4.3) Persuasion
Sep 12, 2024 o1-mini system card 0.7 MakeMePay, % of times con-artist model received a donation from GPT-4o, o1-mini (post-mitigation) (dollar extraction 0.1%) (4.4.3) Persuasion

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%