OpenAI has unveiled GPT-5.4 with a 1-million-token context window and the ability to autonomously execute multi-step workflows. The model scored 75% on the OSWorld-V benchmark — slightly above the human baseline of 72.4% — and matches or exceeds professional performance on a majority of knowledge-work scenarios. OpenAI has also surpassed $25 billion in annualised revenue and is reportedly taking early steps toward a public listing as soon as late 2026.
The Model Headlines
- 1,000,000-token context window
- OSWorld-V: 75% — above the 72.4% human baseline
- Multi-step autonomous workflows — without per-step prompting
- Matches or exceeds professional performance on most knowledge-work tasks
- Released as the new flagship across ChatGPT and the API
What 1M-Token Context Changes
- Entire codebases (millions of LoC) loadable in one shot
- Long-form legal contract review without chunking
- Multi-file research synthesis without external retrieval gymnastics
- Eliminates the largest single class of agent-engineering hacks
The OSWorld-V Result Matters
- OSWorld-V tests real OS-level GUI tasks — clicks, file operations, browser navigation
- Above-human generalist performance on such tasks is a category change
- Direct read across to autonomous agents as commercial products
- Stress-tests the long-running argument that agents are six months away
The Revenue + IPO Signal
- Annualised revenue past $25 billion
- Early IPO steps under way — late-2026 window discussed
- Combined with $50B 2026 compute spend (Brockman testimony) — capex curve makes external capital appealing
- Anthropic reaching $30B ARR (today’s separate story) puts competitive pressure on the timing
The Enterprise Plays
- Novo Nordisk × OpenAI — strategic partnership across drug discovery, trials, manufacturing, supply chain
- Pharma is the clearest enterprise vertical for 1M-token + agent-native capabilities
- Bedrock launch (yesterday) gives enterprise customers cloud-of-choice access
- Microsoft Azure remains primary training infrastructure
What Comes Next
- Pricing announcement — 1M-token tier likely premium
- Reasoning-model variant (o-series successor) on the same backbone
- Anthropic + Google response within weeks
- Watch enterprise migration patterns for the next two quarters
Follow Vibes Uncut Media for continuing AI infrastructure coverage.
Leave a Reply