THE HOOK
GPT-6 Astra beat Valve’s Portal in under 24 hours, then handed everyone a $571.18 invoice for the privilege.
THE GIST
OpenAI’s latest model was used by developer cozyblaze to play through the puzzle game, completing the run in less than a day. The reported cost was $571.18, mostly from the huge amount of computer-use inference required.
THE TAKE
This is a better demo than another benchmark score because it exposes the ugly bit: capable agents burn money while they work. Astra can reason, click, retry and keep going, but every mistake is still a metered API call rather than a bruised ego. The model won the game. Your finance director may not.
OPINION
The Portal run is a neat proof that AI agents can now grind through long, stateful tasks rather than merely produce a clever first move. But The Verge’s report also supplies the important punchline: reliability is meaningless if the economics are allowed to wander off unsupervised.
That matters far beyond games, because a business workflow has no magical finish line where the model politely stops spending. An agent that burns five hundred dollars to solve a toy problem is a demonstration; an agent that does the same thing while reconciling invoices is an incident report. The useful metric is not whether Astra can finish a task, but whether it knows when the task is no longer worth finishing. That requires hard budgets, checkpoints and a human kill switch, not another glossy demo. Otherwise “agentic” just means the meter keeps running after everyone has stopped watching. A benchmark that cannot state its operating budget is just performance art with an API key. Enterprise buyers should demand the cost curve before they applaud the clever finish. Otherwise the demo is merely expensive theatre. That is not a business model. Not remotely. Today.
- Story type: Breaking news
- Impact: High
- Confidence: Medium (multiple reports)
- Audience: Developers
- What changed: GPT-6 Astra completed Valve’s Portal in under 24 hours at a reported cost of $571.18.
- What to watch: Whether OpenAI and developers can reduce the cost of long-running computer-use agents.
- Reading time: 2 minutes
