OpenAI’s o3/o4-mini Models Stir Mixed Reviews and Invisible Marking Debates

April 21, 2025, 7:20 am

OpenAI’s new o3 and o4-mini models are making waves by showcasing impressive coding and math capabilities while paradoxically suffering from increased hallucinations. Adding to the intrigue, observers have discovered the unexpected presence of invisible characters that hint at a built-in watermarking mechanism—because nothing says “cutting-edge” like secret invisible signatures in your output. It’s a curious blend of technical wizardry and quirky oversights that has both experts and skeptics raising an amused eyebrow.

simonwillison.net / OpenAI o3 and o4-mini System Card

OpenAI o3 and o4-mini System Card I'm surprised to see a combined System Card for o3 and o4-mini in the same document - I'd expect to see these covered separately. The opening paragraph calls out the most interesting new ability of these models (see also my notes here). Tool usage isn't new, but...

medianama.com / New OpenAI Models Hallucinating More Than Their Predecessor

OpenAI's new AI models are hallucinating more than their predecessor, according to an internal testing report released by the company. The post New OpenAI Models Hallucinating More Than Their Predecessor appeared first on MEDIANAMA.

simonwillison.net / AI assisted search-based research actually works now

For the past two and a half years the feature I've most wanted from LLMs is the ability to take on search-based research tasks on my behalf. We saw the first glimpses of this back in early 2023, with Perplexity (first launched December 2022, first prompt leak in January 2023) and then the GPT-4...

techspot.com / ChatGPT gets scarily good at guessing photo locations, sparking doxxing concerns

OpenAI released its latest o3 and o4-mini models last week, which can "reason" through uploaded images. This means it can crop, rotate, and zoom in on photos, even if they're of poor quality.Read Entire Article

techspot.com / OpenAI's newest o3 and o4-mini models excel at coding and math – but hallucinate more often

Historically, each new generation of OpenAI's models has delivered incremental improvements in factual accuracy, with hallucination rates dropping as the technology matured. However, internal testing and third-party evaluations now reveal that o3 and o4-mini, both classified as "reasoning models,"...

winbuzzer.com / OpenAI’s New o3/o4-mini Models Add Invisible Characters to Text, Sparking Watermark Debate

The discovery of non-standard space characters in OpenAI's o3/o4-mini output has raised questions about AI watermarking, though it remains unclear if it's intentional. The post OpenAI’s New o3/o4-mini Models Add Invisible Characters to Text, Sparking Watermark Debate appeared first on WinBuzzer.

permalink / 6 stories from 4 sources in 9 hours ago #ai #innovation #ml #dataprivacy #openai #software #analytics #google #anthropic #computervision #techpolicy #openai #artificial-intelligence #ai-ethics #generative-ai #llms

Related Tags

OpenAI’s o3/o4-mini Models Stir Mixed Reviews and Invisible Marking Debates

Related Tags

Artificial Intelligence

Innovation

Machine Learning

Data Privacy

OpenAI

Software

Analytics

Google

Anthropic

Computer Vision

Tech Policy

openai

artificial intelligence

ai-ethics

generative-ai

llms