xAI's budget coding model: agentic-loop pricing for code tasks that don't need a frontier brain. Tiered pricing: $1/$2 per 1M under 200K tokens, $2/$4 per 1M at 200K tokens and above.
Grok Build 0.1 bills $1 / 1M input and $2 / 1M output for prompts under 200K tokens. Once a request reaches 200K tokens, the entire request is billed at $2 / 1M input and $4 / 1M output.
Grok Code Fast 1 has been retired from xAI's official pricing table; Grok Build 0.1 replaces it in the same niche, at the same standard-tier price point, for high-iteration code work. Coding agents burn tokens — every loop re-reads files, re-runs context, retries — and Grok Build 0.1 is priced for that reality: $1 input and $2 output per 1M tokens under 200K means an agent can iterate many times over for a fraction of a frontier-model pass.
It keeps the same 256K window as its predecessor — smaller than flagship Grok 4.5's 500K, but still ample for repo-scale context. Pricing is now tiered rather than flat: requests that cross 200K tokens bill at $2 input / $4 output per 1M for the whole request, so keeping prompts under that mark keeps costs at the cheaper rate.
Competent code generation, edits, and explanations across mainstream languages at high speed. The use case that fits best: scripted or agentic pipelines where the task is well-defined and a retry costs fractions of a cent. Grok Build 0.1 slots into the same tooling and integrations that supported Grok Code Fast 1, so migrating existing pipelines is a model-ID swap rather than a rebuild.
The honest weakness: hard problems. Subtle bugs, architecture, and multi-file refactors with hidden dependencies want Sonnet 5 or better, or xAI's own Grok 4.3. For pure IDE fill-in-the-middle completion, Codestral is the more specialized (and cheaper-output) tool.
| Model | Input / 1M | Output / 1M | Context |
|---|---|---|---|
| Grok Build 0.1 (<200K) | $1 | $2 | 256K |
| Grok Build 0.1 (≥200K) | $2 | $4 | 256K |
| Codestral | $0.30 | $0.90 | 128K |
| Grok 4.3 | $1.25 | $2.50 | 1M |
| Claude Sonnet 5 | $3 | $15 | 1M |
| GPT-5.4 | $2.50 | $15 | 272K |
The direct rival is Codestral: cheaper output and FIM completion, but less agentic framing. xAI's own Grok 4.3 is the step up when reasoning depth matters more than raw throughput. Build 0.1 earns its keep as the volume layer under a smarter model.