Grok 4 5 Prompting

# Grok 4.5 Prompting Guide (Grok Build)

Checked: 2026-07-12

Grok 4.5 is the **default Grok Build model**. SpaceXAI documents **Low,
Medium, and High** reasoning with **High as the default**. Independent
composites place Grok 4.5 High near other frontier coding agents at much lower
published task cost in the dated Grok Build harness results.

| Property | Value |
| --- | --- |
| Role | Cost-efficient coding agent + live web/X search |
| Efforts | Low, Medium, High (default **High**) |
| Base API price (dated) | $2 / $6 per 1M input/output |
| Long-input note | Independent reports of higher rates above ~200K; verify live xAI pricing |
| Harness | Grok Build: repo search, multi-file edit, terminal, tests, Git, recovery, subagents |

## When to Choose Grok 4.5

Choose Grok Build when:

- coding cost and speed matter;
- you want live web or X search in the same agent loop;
- the repository task is well bounded with clear tests.

Validate for your languages, terminal recovery needs, and safety policy. Public
composites can reverse on private workloads.

## Effort Mode Playbooks

### Low

**Best fit:** small mechanical edits, formatting, obvious one-file fixes.