PlainAI AI explained plainly

xAI · 2025 · 07 · 09 · Model · ~1 min read

xAI launched Grok 4

xAI's thinking model upgrade with multi-agent workflows and stronger real-time data integration through DeepSearch. Caught xAI up to where the rest of the leaders had moved during the first half of 2025. Available through X Premium+ and the xAI API.

What's actually new

  • Multi-agent workflows built in. Several Grok instances can collaborate on a task — early version of the agent-team idea other labs would copy through 2025.
  • DeepSearch matured. Live web reasoning got more reliable: better citations, fewer dead links, faster lookups.
  • Bigger context window than Grok 3 — closer to OpenAI and Google's flagship sizes.

If you want more

Worth knowing~30s
  • Marketed as 'top of leaderboards' at launch. Independent comparisons placed it close to GPT-5 (released a month later), not above.
  • Quick-fire upgrades through late 2025 (Grok 4.20, 4.3) made the version landscape confusing for developers.
Who should care~20s

X Premium+ subscribers wondering what they're paying for. Developers comparing thinking models on price-per-call. Anyone curious about whether xAI's positioning produces meaningfully different output from OpenAI's or Anthropic's.

What to do about it~20s

Try Grok 4 through DeepSearch for one real task that needs current information — that's where xAI's 'live' bet pays off. For everything else, pick by your existing tooling.

Honest take~45s

Grok 4 was xAI catching up to the leading edge, not pulling ahead. The multi-agent angle was the more interesting bit — letting several Grok instances tackle one problem together prefigured the agent-team designs Anthropic, OpenAI, and Google all shipped through late 2025. xAI's place in the cosmos by mid-2025 was 'fast follower with looser content rules', which is a real position even if it isn't a winning one.

Other recent model updates

Sources

Last verified · 2026 · 05 · 05 · Found a fact wrong? corrections@aguidetocloud.com