Grok Build 0.1
xAI's dedicated coding model, built for agentic software engineering, code review, and debugging across large repositories.
You always get the exact model you pick — we never silently route you to another.
About Grok Build 0.1
Grok Build 0.1 is xAI's dedicated coding model and the recommended successor to Grok Code Fast 1 — purpose-built for agentic software engineering rather than general chat. Its defining architectural choice is that reasoning is always on: there is no non-reasoning mode and no configurable effort level, so every response works through structured analysis before producing output. That makes it well-matched to multi-step coding loops where deliberate thinking pays off, but it also means the model leans toward tool-calling, MCP-native behavior, which is why it shines inside an agentic harness and needs clear framing when used as a plain conversational assistant. At $1.00 per million input tokens and $2.00 per million output, it lands roughly 20% cheaper than the flagship Grok 4.3 while keeping a 256k-token context window — enough to hold large repositories across multi-step tool calls. Cached input drops to $0.10 per million, and it supports structured outputs, function calling, web search, and prompt caching. The honest caveats: it is text-only (no documented image input), and xAI has not yet published an independent intelligence benchmark, so judge it on real coding tasks rather than a headline score.
Best for
- Agentic coding harnesses — multi-step autonomous edits, tool calls, and MCP-native workflows across large repositories
- Code review and debugging — loading sizeable codebases into the 256k context to trace bugs and propose fixes
- Refactoring and migrations — structured, reasoning-first edits where always-on analysis reduces careless mistakes
- Structured-output tasks — generating typed JSON, function-call arguments, and schema-bound responses for build pipelines
- Cost-sensitive engineering work that wants stronger reasoning than Grok Code Fast 1 without paying flagship Grok 4.3 rates
Specifications
| Provider | xAI |
|---|---|
| Intelligence | High |
| Speed | Fast |
| Context window | 256,000 tokens |
| Input price | $1.00 / 1M tokens |
| Output price | $2.00 / 1M tokens |
| Request cost | 1 premium request |
| Plan tier | Premium |
| Input | Text |
| Output | Text |
| Features | Cached input: $0.10 / 1M tokens, Always-on reasoning (reasoning effort is not configurable), Structured outputs, web search, function calling, and prompt caching supported |
| API endpoints | v1/chat/completions, v1/responses |
| Model ID | grok-build-0.1 |
Frequently asked questions
$1.00 per million input tokens and $2.00 per million output tokens, with cached input at $0.10 per million. That is about 20% cheaper than Grok 4.3 ($1.25 input / $2.50 output).
256,000 tokens — sized for holding large repository context across multi-step agentic tool calls.
No. Reasoning is baked into Grok Build 0.1 and cannot be disabled or tuned. Unlike Grok 4.3 or Grok 3 Mini, it does not accept a reasoning_effort parameter — every call reasons before answering.
No. Grok Build 0.1 is documented as a text-in, text-out model. For image understanding, Grok 4.3 supports image input.
It is the recommended migration target as Grok Code Fast 1 is deprecated. Build 0.1 is the newer dedicated coding model, with always-on reasoning and a higher price point than the speed-first Code Fast 1.
It is trained for MCP-native agentic coding, so it leans toward calling tools. In a plain chat interface it works best when the prompt makes clear it should reply conversationally rather than invoke tools.