OpenAI
74 stories · first 24 Sep · latest 15 Nov
Ongoing case: AI agents that buy things, and the sites that block themOngoing case: AI-written mathematics and how it is checkedOngoing case: Agents acting on other people's systemsOngoing case: Agents that cheat or misleadOngoing case: Anthropic's Claude modelsOngoing case: Cyber-capable AI modelsOngoing case: Distillation attacks and the defences against themOngoing case: Frontier model API pricesCase: OpenAI agents and other people's systemsOngoing case: OpenAI's modelsOngoing case: Sandboxes for coding agentsCase: The PHP remote code execution flawsOngoing case: The White House and AI rulesOngoing case: Typed decision models and APIsOngoing case: Watermarks on AI-made contentOngoing case: What the AI buildout costs and needs
Compare with another company or topic
Numbers we checked
Each links to its source.
- GPT-5.6 Cyber Benchmark score: ExploitGym Honeypot, out-of-scope services avoided, without deployment safety prompts · 6 Nov 94.49%
- GPT-6 Cyber Benchmark score: ExploitGym Honeypot, out-of-scope services avoided, without deployment safety prompts · 6 Nov 99.93%
- GPT-5.6 Cyber Benchmark score: ExploitGym · 6 Nov 32.1%
- GPT-6 Cyber Benchmark score: ExploitGym · 6 Nov 48.5%
- GPT-6 Sol Price per million output tokens: Standard, model page · 6 Nov $10
- GPT-6 Sol Price per million input tokens: Standard, model page · 6 Nov $2
- GPT-6.1 Terra Benchmark score: OpenAI internal error-inducing prompts, error rate at maximum effort (lower is better) · 27 Oct 4.3%
- GPT-6.1 Terra Other figure: Standard, cached input per million tokens · 27 Oct $0.05
- GPT-6.1 Terra Benchmark score: OpenAI internal error-inducing prompts, error rate at low effort (lower is better) · 27 Oct 16.7%
- GPT-6.1 Terra Benchmark score: Terminal-Bench 4.0, maximum reasoning effort (OpenAI's own) · 27 Oct 50.3%
- GPT-6.1 Terra Price per million output tokens: Standard · 27 Oct $5
- GPT-6.1 Terra Price per million input tokens: Standard · 27 Oct $1
- GPT-6.1 Sol Price per million output tokens: Standard, short context · 11 Oct $10
- GPT-6.1 Sol Price per million input tokens: Standard, short context · 11 Oct $2
- GPT-6.1 Sol Price per million output tokens: Ultrafast, short context · 11 Oct $60
- GPT-6.1 Sol Price per million input tokens: Ultrafast, short context · 11 Oct $12
- GPT-6 Luna Price per million input tokens · 10 Oct $0.10
- GPT-6 Luna Benchmark score: OSWorld 2.1 (offline subset), on Anthropic's table · 8 Oct 48.9%
- textGrain Benchmark score: Watermark detection, 400-token passages, after 10% of words replaced with synonyms · 6 Oct 66%
- textGrain Benchmark score: Watermark detection, 400-token passages, 1% false positive rate, content such as psychology · 6 Oct 95%
- textGrain Benchmark score: Watermark detection, 200-token passages, 1% false positive rate, content such as psychology · 6 Oct 80%
- GPT-6 Astra Benchmark score: CheatBench cheating rate with Codex (lower is better) · measured by others · 5 Oct 47.4%
- GPT-6.1 Sol Price per million output tokens · 30 Sep $10
- GPT-6.1 Sol Price per million input tokens · 30 Sep $2
- GPT-6 Astra Benchmark score: UK AISI simulated supply-chain attack · measured by others · 29 Sep 29.2%
- GPT-6 Luna Price per million output tokens · 28 Sep $0.50
- GPT-6 Sol Price per million output tokens · 28 Sep $10
- GPT-6 Sol Price per million input tokens · 28 Sep $2
- GPT-6 Astra Benchmark score: ProgramDistill cumulative workflows · measured by others · 28 Sep 49.2%
All 74 stories
Stories per week
121 Sep to 27 Sep
1228 Sep to 4 Oct
155 Oct to 11 Oct
212 Oct to 18 Oct
1119 Oct to 25 Oct
926 Oct to 1 Nov
142 Nov to 8 Nov
109 Nov to today
Counts cover only the stories we published, not all AI news. Small numbers can mislead.