Allie K. Miller says GPT-5.6 Sol outperforms Claude on computer-use tasks due to reliable command execution
Bindu Reddy adds that the model is faster and cheaper than Opus.
Story Brief
Bindu Reddy adds that the model is faster and cheaper than Opus.
Allie K. Miller@alliekmillerTECH
- OpenAI continually outperforms Anthropic models on computer use (with Claude, I’m surprised when it works, with Codex, I expect it to work) - when I prompt with Fable 5.5, I feel like I’m motivating a “human”, Claude has always had more empathy and better chat vibes, but that impacts how I prompt it. With GPT-5.6 Sol, I just say the thing I want. - one of the use cases my team and I talk about is “battering ram” where an AI attempts every possible move to get something over the line. GPT-5.6 Sol felt more determined in getting over that finish line (I also find GPT models better at idea generation so maybe it’s also more creative in how it can solve it?). If these models are better battering rams, then verifiable goal setting becomes much more important - I have no idea how much 5.6 will cost (if they told us, I didn’t see it), but I suspect it will be less than Fable 5 and that AI budgets everywhere at bleeding edge enterprises will be reviewed in detail (and likely increased for engineering teams) - AI has a massive PR problem on capability and GPT-5.6 and Fable 5 is going to make it worse. I spoke with a woman in Scotland a few days ago. She’s a content designer at a bank in Canada. She’s a techie - she uses Figma, she is masterful with her phone, she knew hardware I barely knew. But she said AI was “just ok”. I asked what she used - it was Gemini mobile app with 3.1 flash model. Imagine engineers wielding Fable 5 while the “standard” access point is 3.1 Flash. AI labs need to solve the model selection crisis ASAP or risk increasing the gap between superusers and standard users and losing the trust of millions. - the AI words of the year are PROACTIVITY and AMBITION and we will see AI systems (groups of models) that can complete a full week or quarter of an entire team’s work. Hiring managers should screen for ambition in candidates, honestly.
