A fast, low-cost model for agentic coding
The successor to the grok-code-fast-1 line; reasons by default and streams a summarized trace.
Overview
Grok Build 0.1 is xAI's fast, low-cost model for agentic coding, released on April 16, 2026. It is the successor to the grok-code-fast-1 line (and still answers to that alias), tuned for the high-volume loop of modern software work — plan, write, test, debug — rather than one-shot answers. Instead of an effort dial, it reasons by default and streams a summarized trace of that reasoning so coding agents can stay steerable.
The design goal is throughput and price: keep an agent moving through many tool calls per task without the latency or cost of a frontier reasoning model. It speaks the OpenAI-compatible API with native function calling and structured outputs, and is strong across TypeScript, Python, Java, Rust, C++, and Go.
Key capabilities
| Dimension | Detail |
|---|---|
| Context window | 256,000 tokens |
| Max output | 256,000 tokens |
| Input modalities | Text, image |
| Output modalities | Text |
| Reasoning | On by default, with streamed reasoning traces (no effort dial) |
| Tools | function calling, structured outputs, streaming, vision |
Inputs above 200K tokens enter a higher long-context tier (2x pricing). Cached input is billed well below fresh input. See live pricing in the model catalog.
The agentic coding loop
The plan → write → test → debug inner loop
High token throughput resolves dozens of tool calls quickly instead of stalling on a slow reasoner.
Grok Build 0.1 is built for the inner loop of a coding agent: read the repository and the task, plan an edit, apply it, run tests or commands, then read the failure and try again. Its high token throughput — xAI reports 100+ tokens/second — is what makes that loop feel interactive, with dozens of tool calls resolving quickly instead of stalling on a slow reasoner. The 256K window keeps large multi-file projects and long histories in a single session, and native function calling plus structured outputs make it a clean fit for editor and CLI agents.
Because it exposes a summarized reasoning trace, tools can surface why the model made a change, which helps developers review and steer long autonomous runs.
When to use it
- Coding agents: editor and CLI agents that fire many tool calls per task.
- High-volume, cost-sensitive work: bulk refactors, test generation, and routine edits.
- Interactive loops: tasks where fast plan-write-test-debug cycles beat peak reasoning.
- Multi-language repositories: TypeScript, Python, Java, Rust, C++, and Go.
CrossModel exposes Grok Build 0.1 through an OpenAI-compatible /v1/chat/completions API. Current pricing is available in the model catalog.