Models
OpenAI
Reduction in AI scheming rates after deliberative alignment

OpenAIX · Sep 17, 2025 · official

Reduction in AI scheming rates after deliberative alignment

OpenAI's benchmark bar chart drew 42.5K views on X: 0.3x its median post and #568 of 1,162 AI chart posts we hold by views. Read Oct 12, 2026.

Reduction in AI scheming rates after deliberative alignment training

XData Benchmark bar chart GPT mini models All OpenAI models o4-mini o3o4-mini after anti-schemingo3 after anti-scheming

OpenAI: Reduction in AI scheming rates after deliberative alignment — Published by OpenAI on X, Sep 17, 2025. Source: x.com

50/ 100

ViralChart score

Reached and resonated

More views and a higher engagement rate than the median post: it travelled and people acted on it.

#618 of 1,162 AI chart posts we hold · #50 of 74 OpenAI posts. Score = views 40%, likes 30%, engagement rate 15%, save rate 15%, each as a percentile of every AI chart post we hold.

Our read

The post sits in the Reached and resonated quadrant at #618 of 1,162 AI chart posts.

Its 0.008 views per follower sits at the 3rd percentile across posts with known followers, in the bottom decile.

At 42.5K views it runs 0.26x OpenAI's 162K median, ranking #58 of 74 OpenAI posts.

Its score of 50 lands at #618 of 1,162 AI chart posts and #50 of 74 OpenAI posts.

0.01views per follower

42.5K views from an account with 5.4M followers.

0.3xOpenAI's median post

42.5K views, where OpenAI's typical post gets 161.5K.

1.0xthe median benchmark bar chart post

Across 215 benchmark bar chart posts we hold, the median gets 44.4K views.

bottom 49%save rate

0.075% of viewers bookmarked it to come back to. The median post gets 0.076%.

Public X numbers from our read on Oct 12, 2026. Peers: 1,162 AI chart posts we hold and 74 from OpenAI.

Where it sits

Reach against resonance

Each dot is a post. Right is more views, up is a higher engagement rate. The lines are the medians, so the four corners are the four kinds of post.

Among 1,162 AI chart posts we hold

Score #618 of 1,162

Views	42.5K	#568	top 49%
Likes	289	#580	top 50%
Engagement rate	0.74%	#577	top 50%
Save rate	0.075%	#589	bottom 49%

Among 74 OpenAI posts

Score #50 of 74

Views	42.5K	#58	bottom 22%
Likes	289	#53	bottom 29%
Engagement rate	0.74%	#23	top 30%
Save rate	0.075%	#15	top 20%

Growth

One read so far

Views42.5K

Age at our read390 days

Views per day of life109views divided by days since it was posted

Dots are our reads. The dotted line from the day it was posted to our first read is a gap, not data.

OpenAI's posts over time

Every post from this account we hold, by date and views. This one is marked; the line is the running median of the last 10.

News · Sep 17, 2025

OpenAI posts anti-scheming benchmark chart for o4-mini and o3

A benchmark bar chart on September 17, 2025 shows scheming rates for o4-mini and o3 before and after anti-scheming training.

What happened

OpenAI published a benchmark bar chart on September 17, 2025 on reduction in AI scheming rates after deliberative alignment training. It was the account's only post that day.

The numbers

It drew 42.5K views, ranking #568 of 1,162 AI chart posts and #58 of 74 OpenAI posts. Its score of 50 ranks #618 of 1,162 AI chart posts.

The lesson

A benchmark bar chart with 4 models shown drew 42.5K views against a 39.9K median among AI chart posts. Its save rate of 0.075% and engagement rate of 0.74% are the measured engagement against saves for this post.

What it means

At #50 of 74 OpenAI posts by score, this chart sits below the account's own median. The 0.26x gap to OpenAI's 162K median views is the measured distance for this publisher's chart posts.

Written by ViralChart from public numbers read on Oct 12, 2026. Every number is checked against our data before it is shown.

Updates: what OpenAI posted next

Data and API

4 points read off this chart

Estimated from the image 97%

Series	Group	Value	Read
o4-mini	o4-mini	8.7 %	printed
o4-mini	o4-mini after anti-scheming	0.3 %	printed
o3	o3	13 %	printed
o3	o3 after anti-scheming	0.4 %	printed

Max plan · API

GET https://viral-chart.com/api/posts/6a75a48ea05e.json
GET https://viral-chart.com/api/charts/6a75a48ea05e.json
GET https://viral-chart.com/api/charts/6a75a48ea05e.skill.md

Make one like this Open the post JSON Chart data JSON Skill file API docs

Static files in this prototype. The post JSON holds the public numbers, ranks, score, quadrant, our reads and, when checked, our copy.

API · Max

Use this chart via API

On the Max plan you can pull this chart's numbers with the provenance of each one, as JSON or CSV, with its redraw spec, or call it from an agent over MCP. Chart id 6a75a48ea05e.

See the Max plan API docs

	Path	Returns	Plan
GET	`/api/v1/charts/6a75a48ea05e`	Title, publisher, date, form, point count; with a Max key, the data link	Open
GET	`/api/v1/charts/6a75a48ea05e/data`	Every point with its provenance and confidence, plus the redraw spec. Add ?format=csv for CSV	Max
GET	`/api/v1/charts/6a75a48ea05e/render?view=1`	The draw-only view this page uses	Open
POST	`/mcp`	Tool get_chart_data with this id, for Claude, Cursor and other agents	Max

curlMax

curl -s https://viral-chart.com/api/v1/charts/6a75a48ea05e/data \
  -H "Authorization: Bearer $VIRALCHART_KEY"

Claude CodeMax

claude mcp add --transport http viralchart https://viral-chart.com/mcp \
  --header "Authorization: Bearer $VIRALCHART_KEY"

60 requests a minute per key; over that you get 429 with Retry-After. Up to 5 keys per account, made on your account page and shown once. Without Max, the data route answers 402 with a link to the plans. The numbers describe a chart its publisher owns, and every file carries a license note.

More charts about GPT mini models

GPT mini models: SWE-Bench Verified Software Engineering

OfficialX · OpenAIApr 16, 2025

SWE-Bench Verified Software Engineering392.1K views · 707 likes

Data

GPT mini models: BrowseComp Agentic Browsing

OfficialX · OpenAIApr 24, 2025

BrowseComp Agentic Browsing367.2K views · 1.5K likes

Data

GPT mini models: Competition math AI model performance

OfficialX · OpenAIAug 5, 2025

Competition math AI model performance146.6K views · 722 likes

Data

GPT mini models: Codeforces competition code Elo rating

OfficialX · OpenAIAug 5, 2025

Codeforces competition code Elo rating146.6K views · 722 likes

Data

GPT mini models: Humanity's Last Exam Expert-level questions across subjects

OfficialX · OpenAIAug 5, 2025

Humanity's Last Exam Expert-level questions across subjects146.6K views · 722 likes

Data

OfficialBlog · OpenAIApr 16, 2025

Trend Comparison

More from OpenAI

OpenAI line chart: Output tokens by department

2.3x

OpenAI·Jun 2026

Output tokens by department

1.3M views5.4K likes0.5% eng.

Data

OpenAI annotated chart: Research paper benchmarking

2.9x

OpenAI·Apr 2025

Research paper benchmarking

1.1M views7K likes0.8% eng.

2.0x

OpenAI·Sep 2026

AI model performance

999.9K views4.8K likes0.6% eng.

Data

ChatGPT: Artificial Analysis Intelligence Index v4.1

OfficialX · OpenAIJul 30, 2026

Artificial Analysis Intelligence Index v4.121M views · 19.5K likes

Data

OfficialX · OpenAIMar 5, 2026

AI Model Benchmark Comparison7.2M views · 23.2K likes

Data

ChatGPT: GPT4 Turbo Benchmark Comparison

OfficialX · OpenAIApr 12, 2024

GPT4 Turbo Benchmark Comparison6.4M views · 6.7K likes

Data

OfficialX · OpenAISep 3, 2026

AI Model Benchmark Comparison5.4M views · 10.3K likes

Data

OfficialX · OpenAIDec 11, 2025

AI Model Benchmark Comparison2M views · 3.8K likes

Data

GPT-5.6: TerminalBench 2.1 AI Model Scores

OfficialX · OpenAIJun 26, 2026

TerminalBench 2.1 AI Model Scores1.9M views · 3.9K likes

Data

Questions people ask

Where does this chart come from?

OpenAI published it on Sep 17, 2025 on X (x.com). We keep it here, with the publisher and date credited as text, so you can study it without leaving.

Can I make a chart like this?

Use the format, not the picture. Open its data page and use Make one like this: it starts from this chart's numbers and takes yours.