Grok and Fable 5 are the most predictable frontier models. Opus and Sol vary a lot between runs.
If you're building automations, remember this.
Poland, Katowice
Joined July 2026
- Grok spent the most tokens and did the most searches out of all the frontier LLMs to complete the task Despite being more thorough and less efficient, it is still cheaper than Opus and Fable and comparable to Sol
- We're now onboarding multiple people a day Out of all the models, Opus is the one that fails most frequently at simply installing DeepAPI, often outright refusing to do soOpus 5 is out of control it does not listen to instructions at all...



