OpenAI Launches GPT-5.4: One Model for 40+ Professions – Replacing Separate Coding Models

OpenAI has rolled out GPT-5.4, its latest frontier model, across ChatGPT, Codex, and the API. The promise is clear: one model that handles complex work across more than 40 professions.

What's New?

GPT-5.4 launches in two variants: standard and GPT-5.4 Pro. The model is designed to eliminate the need for separate specialized models in most use cases – including coding.

Key metrics:

• OSWorld-Verified: 75% task success rate (vs GPT-5.2's 47.3% and 72.4% human baseline)
• Context window: 1.05 million tokens (long-context) / 272,000 tokens (standard)
• Token usage: New search method cuts token consumption by nearly half in large systems
• Stronger performance on WebArena-Verified and Online-Mind2Web – better for automated web workflows

Desktop Control and Agent Capabilities

GPT-5.4 can operate desktops and web browsers via screenshots, mouse input, and keyboard commands. This positions the model as a real tool for agentic desktop automation – not just chat.

Professional Coverage

The model covers finance, law, engineering, and more. OpenAI claims GPT-5.4 produces fewer false claims than GPT-5.2, and now provides more visibility into its reasoning steps through the system card.

Relevance for CIOs

For enterprises evaluating AI investment, GPT-5.4 reduces operational complexity by consolidating functionality that previously required multiple models. Combined with near-halving of token costs in large systems, this can deliver concrete ROI in enterprise deployments.