OpenAI released its o3 and o4-mini models that can reason and use ChatGPT's multimodal tools. The release of its latest models are designed to tackle multi-faceted questions more effectively, a step toward a more agentic ChatGPT."
Ultimately, OpenAI is planning to roll up its models to enable ChatGPT to complete tasks for you. OpenAI is billing its latest models as "the smartest models we've released to date."
The launches come just days after OpenAI launched ChatGPT 4.1. Perhaps the biggest takeaway (aside from OpenAI's model naming conventions are confusing) is that ChatGPT will be more agentic and ultimately abstract the various models underneath.
Latest LLM news:
- Google Gemini vs. OpenAI, DeepSeek vs. Qwen: What we're learning from model wars
- Anthropic Claude Research integrates with Google Workspace
- Cohere rolls out Embed 4, an enterprise multimodal search model
According to OpenAI, o3 is its most powerful reasoning model and excels at coding, math, science and visual perception. OpenAI's o4-mini is a smaller model that's optimized for fast reasoning at a lower cost.
Here's what you need to know about the latest release:
- The new models can integrate images into their chain of thought and can handle whiteboard photos, textbook diagrams and sketches.
- OpenAI o3 has equal latency and cost with OpenAI o1 with higher performance.
- Both models can deploy tools through reinforcement learning when there's something it doesn't know.
- ChatGPT Plus, Pro and Team users will see o3, o4-mini, and o4-mini-high in the model selector to replace o1, o3‑mini, and o3‑mini‑high.
- Enterprise and education accounts get access in a week.
- OpenAI o3-pro will be released next.
Separately, OpenAI announced Codex CLI, which is a lightweight coding agent that can run from terminals. It works on the computer and id designed for o3 and o4-mini with support coming for GPT-4.1.
