At a glance
WHAT IT’S REALLY ABOUT
GPT-5.6 Sol beats Claude Fable on Claire’s real-world benchmark
- OpenAI’s GPT-5.6 family is framed as three tiers—Sol (frontier), Terra (balanced), and Luna (cheap/high-volume)—with Sol positioned as Claire’s preferred daily driver.
- Claire’s “How I AI” benchmark blends an LLM judge (GPT-5.5) with her own taste test (70% Claire / 30% machine), and under that weighting GPT-5.6 Sol wins overall by a meaningful margin.
- Sol is praised for producing more functional, opinionated, and less “template-slop” prototypes than Fable, especially on dense/technical UIs like ops dashboards and incident triage tools.
- Fable is credited as highly intelligent and capable for complex technical work, but Claire finds it overly pedantic and hard to collaborate with due to unnatural, overly-engineered communication.
- Beyond the benchmark, Claire highlights Sol’s practical effectiveness in breaking through overly rigid architectures, plus strong workflows for video clipping/editing and high-throughput browser automation via Codex + @Chrome.
IDEAS WORTH REMEMBERING
5 ideasSol wins because it’s practically shippable, not just theoretically smart.
Claire’s core critique is that Fable can be “hyper-intelligent” yet overly constrained and pedantic, while Sol more readily loosens constraints to deliver usable product outputs aligned with end-user goals.
Pricing favors Sol, making it easier to standardize on for frequent work.
She cites Sol at ~$5/M input and ~$30/M output tokens versus Fable at ~$10/M input and ~$50/M output, arguing the cost gap matters when prototyping and iterating heavily.
Functionality inside prototypes is a primary differentiator in real workflows.
In side-by-sides (ops dashboard, creative pack site, incident triage, habit trackers), Sol more often produced interfaces where expected interactions actually worked and the layout supported real task flow.
Terra can be the better pick for straightforward business writing and PRDs.
Despite Sol’s overall win, Claire preferred Terra’s PRD outputs as “streamlined and to the point,” implying Terra may be optimal when clarity and brevity beat flourish.
Sonnet remains Claire’s favorite for “agentic voice,” even if it’s imperfect.
For assistant-like interactions (rescheduling meetings, handling deploy stress), she rates Sonnet as most human-sounding, while Sol (and especially Fable) is criticized for cringe phrasing and em-dash-heavy delivery.
WORDS WORTH SAVING
5 quotesI did blind taste test these, and so I do really feel like it did a good job.
— Claire Vo
I hate slop. I hate slop. I hate slop. We all hate slop. It's the worst. It's the worst part of AI.
— Claire Vo
I hate talking to Fable 5 because it talks to me like an engineer that has never met a human before. It's like its first day on Earth.
— Claire Vo
If you would take away, like, one highlight, um, difference between Fable and Sol is, like, Fable is theoretically hyper-intelligent and Sol is practically effective.
— Claire Vo
The problem is when you're building products, exact precision is neither helpful nor possible.
— Claire Vo
High quality AI-generated summary created from speaker-labeled transcript.
