Three days. That’s the gap between grok-build-0.1 hitting the public API (May 29) and Composer 2.5 landing in the CLI (June 1). xAI’s blog confirms both events directly, “Grok Build 0.1 on API” on May 29, and “Composer 2.5 is now available in Grok Build. Try it from the /models menu” on June 1. No ambiguity on dates, no vendor narrative, just a release log that tells you something about how xAI is operating right now.
The prior coverage (published May 31 on this hub) covered the grok-build-0.1 API beta and its core specs. This brief is about what’s new: Composer 2.5.
Composer 2.5 is xAI’s iteration on the composer pattern familiar to developers using Cursor or GitHub Copilot’s agent mode. The tool lives inside the Grok Build CLI, accessible from the /models menu. xAI hasn’t published detailed capability comparisons or independent benchmarks for Composer 2.5 specifically, what’s confirmed is the launch date, the platform it runs on, and the fact that it’s available now.
Agentic Coding API Pricing (vendor-reported where noted)
The underlying model specs: xAI says grok-build-0.1 runs at over 100 tokens per second, optimized for low-latency agentic loops, and is priced at $1 per million input tokens and $2 per million output tokens. Those are vendor-reported figures, throughput at production scale with real agentic workflows isn’t the same as peak throughput in controlled conditions. xAI says the model is natively trained for agentic coding, automated debugging, web development, and Model Context Protocol workflows. Don’t expect those capability claims to hold uniformly before you’ve run it against your own stack.
The corporate context. xAI, which now operates under the SpaceXAI brand following a reported restructuring under SpaceX, confirmed the “SpaceXAI” entity name in an official May 6 blog post announcing a compute partnership with Anthropic. The characterization of the restructuring as a formal SpaceX division comes from lower-tier sources, the “SpaceXAI” branding is confirmed; the corporate structure specifics aren’t. Worth knowing for vendor evaluations where legal entity matters.
What to Watch
What to watch
The competitive signal here isn’t Composer 2.5 specifically, it’s the release cadence. GitHub Copilot, Cursor, and Codex are all in the same market and all iterating fast. The teams evaluating which agentic coding platform to standardize on should be tracking whether xAI can sustain this pace and whether independent evaluations of grok-build-0.1 emerge. Right now, all performance data comes from xAI.
TJS synthesis: Composer 2.5 is a competitive move in a market where pacing is itself a signal. The $1/$2 pricing is competitive on paper, but without independent throughput benchmarks at production agentic workload levels, the real cost per task is unknown. Wait for third-party evaluations of grok-build-0.1’s actual agentic performance before standardizing on this stack. The CLI is available now; the evidence base for enterprise adoption isn’t.