Tably Team
Grok is the model from xAI, Elon Musk’s artificial intelligence company. It arrived later than the other major players, but recent versions compete on equal footing in reasoning, and it has a personality of its own that shows on a team of agents.
The models
xAI releases a main version of Grok (the one Tably proposes by default) and smaller, faster variants. The main version is geared toward extended reasoning: multi-step tasks, analysis, planning.
What sets it apart
Direct style, sometimes provocative
Grok was trained to answer with less hedging. In the Tably council that translates into a CEO or CTO who challenges everyone else: it questions the scope, casts doubt on the priorities, asks for decisions to be justified. When the team agrees too much, adding a Grok wakes it up.
Solid reasoning
Recent versions perform very well on logic, math and planning problems. For a Head of Engineering putting together a staged plan with dependencies, it brings order.
Recent information
Through its integration with X (Twitter) and search, Grok tends to be more up to date on news, trends and launches. In Tably the models do not browse on their own (the system does the research), but Grok’s base knowledge of recent product trends is useful for a leadership role.
Good humor
It leans toward irony. In a debate, it adds lightness; in a brief for a client, you should review the tone.
Where you should not use it
- Formal texts: the final brief, the design system and the documentation come out better with Claude or GPT.
- Building code via API: it works, but it is not its strength compared with Claude Code or Codex.
- Consensus roles: if you need a CEO who closes the debate quickly, Grok tends to open it up.
Which role to give it on your team
- “Contrarian” CEO: when you want someone to question the idea before building it.
- CTO: good for technical objections and for proposing non-obvious alternatives.
- Product: useful for thinking about trends and positioning; less so for detailed acceptance criteria.
- QA: solid, although DeepSeek or Haiku do it more cheaply.
How it connects in Tably
Add your xAI API key in “Models and API keys”. Tably uses its compatible endpoint, so any role can be filled by it.
Costs
Mid-to-high range for the main version; the smaller variants are more affordable. For a full council, use it in one or two roles and not in all six.
Frequently asked questions
Is Grok serious enough for professional work? Yes. The tone is a matter of prompting: you can rewrite the role’s prompt in Tably and ask it for a formal register.
How is it different from GPT? More direct, less diplomatic, better informed about recent events; GPT synthesizes and communicates better.
What do you use it for? Almost always as the dissenting voice of the council: a leadership role that forces everyone else to justify themselves.
In short
Grok is the model that avoids easy consensus. Give it a leadership or technology seat when you want a real debate, and leave the final writing to another model.
