Model

Grok Build 0.1

CLOSED

SpaceXAI · Grok Build family · released May 20, 2026

Confirmed real and distinct: xAI's (SpaceXAI's) first model purpose-built for agentic coding, powering the Grok Build CLI/terminal agent. Opened to developers via the API in public beta on ~May 29, 2026, a bit after the underlying model went live (~May 20, 2026). 256K context, text+image input, always-on reasoning, function calling, structured outputs, native MCP ('bring your own MCP') support. Not to be confused with 'Grok Build' the product/CLI/agent name, which Grok 4.5 now also powers as of July 2026 — Grok Build 0.1 was the original dedicated coding model before Grok 4.5 became the CLI's default.

ReasoningCodingVisionFunction callingTool useAgentic

2268.3

Elo · rank #37

Parameters

Undisclosed

Active params

—

Context

256K tokens

Architecture

Architecture undisclosed; a dedicated, fast agentic-coding model (100+ tok/s) purpose-built to power the Grok Build CLI, distinct from the general-purpose Grok line

License

Proprietary

Languages

—

API price (in/out)

$1 / $2

Modalities

text · vision

Benchmark results

Bar shows position within the tracked field; marker = field best

GPQA DiamondReasoning89.0%#31

best: GPT-5.6 · 94.6%

Humanity's Last ExamReasoning36.0%#32

best: Claude Sonnet 5 · 57.4%

MMMU-ProVision77.0%#18

best: Claude Opus 4.7 · 85.5%

τ³-BankingAgents11.0%#2

best: Grok 4.5 · 33.0%

Terminal-Bench 2.0Coding52.0%#44

best: GPT-5.6 · 88.8%

Run it locally

Closed weights — available via API only. No local deployment.

Input / M tok

Output / M tok

Grok Build family

Elo progression across releases

Grok Build 0.1May 20262268.3

Resources

DOCShttps://docs.x.ai/developers/models/grok-build-0.1 APIAvailable via SpaceXAI API

API price $1/$2 · each benchmark row carries its own source badge (see methodology)