100+ Tokens per Second
xAI reports serving above 100 tokens per second, reducing wait time across repeated plan, tool, edit, and test cycles.
Grok Build 0.1
Access xAI's coding model through GPTProto for web development, debugging, MCP workflows, and tool-driven agents. Get 100+ tokens-per-second serving, a 256K context window, text and image input, and one API key shared across 200+ models.
Prezzi Grok Build 0.1
Chat, agenti di coding e lavoro sui documenti. Prezzi per 1M di token: input, input memorizzato nella cache e output vengono fatturati separatamente. GPTProto costa il 40% in meno rispetto alle tariffe ufficiali.
| Scenario | Listino ufficiale Grok | OpenRouter | GPTProto | Risparmi al mese |
|---|---|---|---|---|
| Personale10M token/mese (4.8M nella cache) | $8.16 | $8.61 | $4.90 | −$3.26≈ $39.17 / anno |
| Team100M token/mese (48M nella cache) | $81.60 | $86.09 | $48.96 | −$32.64≈ $391.68 / anno |
| Aziendale500M token/mese (240M nella cache) | $408.00 | $430.44 | $244.80 | −$163.20≈ $1958.40 / anno |
Risparmia $0.00 / mese
Risparmio annuale ≈ $0.00 · pagamento a consumo
Modelli correlati
Tutti i modelli
Grok Build 0.1Attuale
Grok · $0.6/M input · $1.2/M output
Bytedance · 262K context · $0.0971/M input · $0.9714/M output
Bytedance · 262K context · $0.0971/M input · $0.9714/M output
Modello
Prezzo: GPTProto
vs ufficiale
vs OpenRouter
Contesto
Stabilità
Grok Build 0.1Attuale
Grok
$0.60 / $1.20per 1M di token
-40%
-43%
—
Grok Build 0.1 API for Fast Agentic Coding
Use Grok Build 0.1 for coding loops that inspect context, call tools, edit files, and verify results. The model is tuned for agentic software work rather than one-shot code completion, with 100+ tokens-per-second serving for faster iteration.
Built for Agentic Coding
Trained for web development, debugging, MCP support, and multi-step software tasks that depend on tools and verification.
256K Context with Image Input
Combine code, logs, specifications, and screenshots within a 256K-token context window, then return text or structured results.
Tool Calls and Structured Output
Connect the model to functions and external systems, or request structured output for plans, patches, checks, and downstream automation.
What Is the Grok Build 0.1 API?
Grok Build 0.1 is a coding model trained for agentic software engineering. xAI announced API access on May 29, 2026 and described it as its fastest coding model, served at more than 100 tokens per second. Its target workloads include web development, debugging, MCP-enabled tools, and workflows with repeated model and tool turns.
It accepts text and image input and returns text. Developers can use it as a text-to-text API for code generation, review, and debugging, or add screenshots as visual references. The model supports function calling, structured outputs, and reasoning within a 256K-token context window.
On GPTProto, the same account, API key, and balance also cover 200+ other models. This makes side-by-side evaluation and fallback routing easier without adding another provider account.
| Specification | Grok Build 0.1 |
|---|---|
| Provider | xAI / SpaceX Super Intelligence |
| API launch | May 29, 2026; announced in public beta |
| Model ID | grok-build-0.1 |
| Input modalities | Text and image |
| Output modality | Text |
| Context window | 256,000 tokens |
| Published serving speed | 100+ tokens per second |
| Core capabilities | Reasoning, function calling, structured outputs |
| Best-fit workloads | Agentic coding, web development, debugging, MCP and tool-calling workflows |
| Standard provider rate | $1.00/M input, $0.20/M cached input, $2.00/M output |
| Long-context threshold | Provider pricing increases when a request exceeds 200K context |
| Batch API | Not supported |
Grok Build 0.1 Model vs Grok Build
Grok Build 0.1 is the API model. Grok Build is the coding-agent application that assembles context, exposes file and command tools, dispatches tool calls, and displays changes. The model alone cannot read a repository, edit files, or run tests; the surrounding agent must provide those tools and continue the loop.
Grok Build 0.1 API Applications
Repository coding agents: Retrieve relevant files, expose edit and test tools, and keep the loop running until acceptance checks pass.
Frontend development: Revise components, use screenshots as visual input, and verify layout with browser or test tools.
Debugging and CI repair: Supply an error trace, related files, recent diffs, and a test command; require reproduction and a rerun after the patch.
MCP workflows: Connect approved repository, issue-tracker, or documentation tools with narrow permissions and validated side effects.
General tool-calling agents: Use fast iterations for structured research, transformation, or automation that fits within 256K tokens.
Grok Build 0.1 API Parameters for Agentic Work
Set the model ID, input or message history, tool definitions, streaming, structured-output schema, and output limit deliberately. Use grok-build-0.1, send only task-relevant context, preserve tool calls and results in order, and validate structured output before execution. Exact field names differ by endpoint, so use the GPTProto API Usage tab as the parameter reference.
Grok Build 0.1 vs GPT-6.1 Sol, DeepSeek V4.1 Flash, Gemini 4 Argon, and Claude Sonnet 5.5
Choose by iteration speed, context length, modality, reasoning control, and cost. Run the same repository task and tool harness before moving traffic.
| Model | Documented Context | Input | Practical Fit |
|---|---|---|---|
| Grok Build 0.1 | 256K | Text, image | Fast agentic coding loops, web development, debugging, MCP, and structured tool use |
| GPT-6.1 Sol | 1.05M | Text, image, document | Complex coding, computer use, and professional work that benefits from adjustable reasoning and much larger context |
| DeepSeek V4.1 Flash | 1M | Text, image | Long-context coding and agent work, switchable thinking modes, and unusually long output requirements |
| Gemini 4 Argon | Up to 1M combined | Text, image, audio, video | Multimodal long-horizon work where audio or video input matters |
| Claude Sonnet 5.5 | 1M | Text, image, document | Coding, document work, and collaborative iteration with a fast hybrid-reasoning model |
| GPT-6.1 Terra | 1.05M | Text, image, document | Coding, visual reasoning, and computer use with adjustable reasoning and a larger context window |
Grok Build 0.1 vs GPT-6.1 Sol: Choose Grok Build for a fast coding loop within 256K. Choose Sol for much larger context, adjustable reasoning, computer use, or broader professional work.
Grok Build 0.1 vs DeepSeek V4.1 Flash: Grok Build is the speed-led coding option. DeepSeek adds a 1M window, thinking and non-thinking modes, and up to 384K output for larger inputs or artifacts.
Grok Build 0.1 vs Gemini 4 Argon: Use Grok Build for focused code agents. Use Argon for audio or video input, broader multimodal reasoning, or close to one million tokens of combined context.
Grok Build 0.1 vs Claude Sonnet 5.5: Grok Build prioritizes coding-agent speed. Sonnet offers a 1M window for coding, knowledge work, and collaborative iteration.
For a price-first comparison across the catalog, use GPTProto's cheapest LLM API directory and confirm the live rate before routing traffic.
Upgrading from Grok Code Fast 1 to Grok Build 0.1
xAI names Grok Build 0.1 as the replacement for grok-code-fast-1. Use the new model ID explicitly instead of relying on an old alias.
Before switching production traffic:
Change the model field to
grok-build-0.1and confirm that the GPTProto API Usage example uses the same ID.Retest tool-call parsing, structured outputs, streaming, and error recovery.
Recalculate the context budget; provider pricing changes above 200K context.
Remove workflows that assume Batch API support.
Run acceptance tests for patch validity, latency, token use, and rollback behavior.
When Should You Choose Grok Build 0.1?
Choose it for fast coding and tool turns when the relevant context fits within 256K tokens and the agent can verify its work with tests or browser checks.
Choose another model for near-million-token context, audio or video input, configurable reasoning depth, or Batch processing. GPT-6.1 Terra is a broader coding and computer-use option, while Gemini 4 Argon covers more input modalities.
Grok Build 0.1 API: Common Technical Questions
How do I get a Grok Build 0.1 API key on GPTProto?
Create an account, add balance, and generate a key from the GPTProto dashboard. The same key accesses this model and 200+ others. Use the model ID shown in the API Usage tab.
What is the Grok Build 0.1 API context window?
The documented window is 256,000 tokens. Retrieve relevant files, summarize older tool results, and compact long histories instead of resending an entire repository. Provider rates increase above 200K context.
Is Grok Build 0.1 only a text-to-text API?
No. It returns text but accepts text and image input, so an agent can combine code or logs with a screenshot. It does not generate images.
How much does the Grok Build 0.1 API cost on GPTProto?
GPTProto currently uses the original provider rate, not a 40% discount. The official standard rate is $1.00/M input, $0.20/M cached input, and $2.00/M output. Check the live panel for cached and long-context billing.
Which Grok Build 0.1 API parameters matter most?
Start with the model ID, input or messages, tools, streaming, structured-output schema, and output limit. Preserve tool calls and results in order, and check the API Usage tab for exact fields.
Can I use Grok Build 0.1 for long-running agent tasks?
Yes, if the agent manages context and verifies each step. Keep relevant code and recent tool results active, return tool errors, and define completion checks. Use retrieval or a 1M-context model when 256K is insufficient.
Does Grok Build 0.1 support function calling and structured output?
Yes. The official page lists function calling, structured outputs, and reasoning. The surrounding agent must still validate arguments, enforce permissions, execute tools, and return results.
Can I try Grok Build 0.1 in an API playground?
Yes. Use the page's Playground button, then repeat the test with the same tool schemas, repository context, and acceptance criteria planned for deployment.
Is Grok Build 0.1 API access private?
Access is authenticated, but that does not imply a dedicated single-tenant deployment. Do not claim zero retention or private hosting unless the current GPTProto policy explicitly confirms it.
Is Grok Build 0.1 the upgrade path from grok-code-fast-1?
Yes. xAI recommends grok-build-0.1 for former grok-code-fast-1 code workloads. Use the new ID explicitly and rerun tool, context, latency, cost, and patch-quality tests.
Articoli correlati
Guide, confronti e aggiornamenti relativi a questo modello.
Tutti gli articoli
Best All-in-One AI APIs for LLMs, Image Generation, and Video: 11 Compared
Compare 11 all-in-one AI API platforms for language, image, video, and audio workflows. Check current modality coverage, billing units, free-access scope, and API formats.
Which AI Model Is Best for SillyTavern in 2026
Compare 8 top AI models for SillyTavern roleplay, long-term memory, local use and lower costs—including Kimi, Gemini, Claude, GPT, GLM and DeepSeek.
Best LLMs for Tool Calling in 2026: 9 Models Compared
Compare nine function-calling LLMs by documented tool support, API restrictions, pricing, and agent fit. Separate valid JSON from correct, successful tool execution.
Best AI for Roleplay in 2026: 6 Models Compared for Memory, Stories, and RPGs
Compare the best AI models for roleplay in 2026 for memory, story writing, RPGs and long chats, including Kimi K3, GLM, Claude, DeepSeek and Gemini.