Grok Build 0.1

What is Grok Build 0.1?#

Grok Build 0.1 is xAI's coding model, the one behind the Grok Build terminal agent, released to the API in early access on May 19, 2026 and as a public beta on May 29. It is trained specifically for agentic coding: web development, debugging, and working through tools, including MCP servers. It accepts text and image input, outputs text, and runs on Unifically as xai/grok-build-0.1 with a 256k-token context window. It is the successor of grok-code-fast-1; xAI now resolves that ID to this model.

Key features of Grok Build 0.1#

First token in half a second

Artificial Analysis measures 0.53 seconds to first token, against a 2.99-second median for reasoning models in its price class, with an Intelligence Index of 41 (#73 of 195) and $0.23 per index task (#15 of 195). It starts answering while other reasoning models are still thinking.

Built for agent harnesses

The same weights power Grok Build's terminal agent, and xAI positions the API model for harnesses such as Cursor, Kilo Code, OpenCode, OpenClaw, and Hermes Agent. Expect it to behave best inside a tool loop, not as a one-shot chat model.

Function calling, structured outputs, and MCP

Native tool calls and JSON-schema outputs through the standard request, plus training on MCP-style tool use, so it slots into existing agent frameworks without prompt gymnastics.

Cheapest xAI coding model per task

At $0.23 per Intelligence Index task it costs about half of Grok 4.5 on the same evaluation, which is the trade for the lower ceiling.

Best for#

Use cases#

Use Grok Build 0.1 as the fast model inside a coding agent: let it read files, run commands, and apply small fixes, then escalate the hard cases to Grok 4.5 or Grok 4.6. It suits web-app scaffolding from a single prompt, automated debugging runs that chew through logs and screenshots, and MCP-driven workflows where the model has to call custom tools many times. Because first tokens arrive in about half a second, it also works for inline code assistance and command suggestion, where a ten-second pause would break the flow. Keep tasks inside the 256k context or add retrieval for bigger repos.

Limitations#

Grok Build 0.1 is a fast model, not a frontier one. Its Intelligence Index of 41 trails Grok 4.5 at 56 and Grok 4.6 at 61, and it is not on the Terminal-Bench 2.1 or SWE-bench Verified boards, so plan for hand-off on hard problems.

Speed is about latency, not throughput: xAI quotes over 100 tokens per second, but Artificial Analysis measures 68.9, just under the 70 median. It is also very verbose, using 130M output tokens across the Intelligence Index against a 71M median, which raises the bill on long tasks.

The 256k context window is the smallest in the current Grok family, and the model is still labeled a beta by xAI.

Grok Build 0.1 vs Grok 4.5#

Grok 4.5 is the stronger coder: 56 against 41 on the Intelligence Index, fourth on Terminal-Bench 2.1, and 86.6% on SWE-bench Verified. Grok Build 0.1 answers faster where it counts, 0.53 seconds to first token against about 10 seconds for Grok 4.5 at high effort, and costs about half per task. xAI itself now defaults Grok Build to Grok 4.5, which tells you where the quality ceiling is. Use Grok Build 0.1 for the fast inner loop and Grok 4.5 for the hard steps.

When to use Grok Build 0.1#

Use Grok Build 0.1 when an agent needs many quick tool calls, when latency shapes the user experience, or when you run large volumes of routine coding work on a budget. Skip it for difficult multi-file changes, long design tasks, or anything where a single high-quality answer beats ten fast ones; that is Grok 4.5 or Grok 4.6 territory.

API examples#

Call Grok Build 0.1 from any language by POSTing to /v1/chat/completions, the OpenAI-compatible endpoint shared by every language model on the platform. Full parameter docs live at docs.unifically.com/models/llm/xai/grok-build-0.1.

curl -X POST https://api.unifically.com/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d '{
    "model": "xai/grok-build-0.1",
    "messages": [
      { "role": "user", "content": "Generate a clean, minimalist personal website with elegant typography as a single HTML file." }
    ]
  }'

The response comes back synchronously with the completion. Set "stream": true to receive tokens as they generate, and pass tools to let it call your functions.