Grok 3 Beta
Grok 3 Beta is xAI's full-scale Grok 3 reasoning model. Trained on the Colossus supercomputer, it targets math, science, and coding benchmarks with a context window of 131.1K tokens.
import { streamText } from 'ai'
const result = streamText({ model: 'xai/grok-3', prompt: 'Why is the sky blue?'})Playground
Try out Grok 3 Beta by xAI. Usage is billed to your team at API rates. Free users (those who haven't made a payment) get $5 of credits every 30 days.
About Grok 3 Beta
Grok 3 Beta was released February 17, 2025 as xAI's full-scale Grok 3 reasoning model, trained on the Colossus supercomputer cluster. It's a step up from the Grok 2 generation on reasoning-heavy tasks across mathematics, science, and code.
Grok 3 Beta supports a context window of 131.1K tokens and up to 131.1K tokens per response. It reports scores on benchmarks including AIME, MATH-500, and GPQA. The model handles complex multi-step reasoning, extended analysis, and nuanced instruction following. It's designed for tasks where depth of reasoning matters more than raw response speed.
The model is accessible through Vercel AI Gateway at $3.0 per million input tokens and $15.0 per million output tokens. For latency-sensitive applications that can tolerate some quality tradeoff, the Grok 3 Fast variant offers faster inference at the same context window size.
Providers
Route requests across multiple providers. Copy a provider slug to set your preference. Visit the docs for more info. Using a provider means you agree to their terms, listed under Legal.
P50 throughput on live AI Gateway traffic, in tokens per second (TPS). Visit the docs for more info.
P50 time to first token (TTFT) on live AI Gateway traffic, in milliseconds. View the docs for more info.
Direct request success rate on AI Gateway and per-provider. Visit the docs for more info.
More models by xAI
| Model |
|---|
| 1M | 0.8s | 94tps | $1.25/M | $2.50/M | Read: $0.2/M Write: — | $5/K + input costs | — |  | 04/30/2026 | ||||
| 2M | 3.8s | 820tps | $1.25/M | $2.50/M | Read: $0.2/M Write: — | $5/K + input costs | — | | 03/11/2026 | ||||
| 2M | 0.6s | 58tps | $0.20/M | $0.50/M | Read: $0.05/M Write: — | $5/K + input costs | — | | 09/19/2025 | ||||
| 256K | 0.3s | 92tps | $0.20/M | $1.50/M | Read:$0.02/M Write:— | — | — | | 08/28/2025 | ||||
| 2M | 0.2s | 207tps | $0.20/M | $0.50/M | Read: $0.05/M Write: — | $5/K + input costs | — |  | 07/09/2025 | ||||
| 2M | 0.6s | 161tps | $0.20/M | $0.50/M | Read: $0.05/M Write: — | $5/K + input costs | — | | 07/09/2025 |
What To Consider When Choosing a Provider
- Configuration: Grok 3 Beta prioritizes reasoning depth over speed. For interactive applications where response time is critical, evaluate whether Grok 3 Fast provides sufficient quality at lower latency.
- Configuration: Deep reasoning tasks can produce lengthy responses. Budget output tokens generously and monitor usage patterns during development.
- Zero Data Retention: AI Gateway does not currently support Zero Data Retention for this model. See the documentation for models that support ZDR.
- Authentication: AI Gateway authenticates requests using an API key or OIDC token. You do not need to manage provider credentials directly.
When to Use Grok 3 Beta
Best For
- Complex mathematical reasoning: Competition-level problems, proofs, and multi-step quantitative analysis
- Scientific research assistance: The model needs to reason across multiple concepts, synthesize findings, and provide detailed explanations
- Advanced code generation and debugging: That benefits from careful step-by-step reasoning through complex logic
- Long-form analytical writing: Research summaries, technical documentation, and detailed comparisons
- Multi-step problem solving: In domains like finance, engineering, and data analysis where accuracy outweighs speed
Consider Alternatives When
- Latency-sensitive applications: Grok 3 Fast delivers comparable results at faster inference speeds
- Simple classification or extraction tasks: A smaller, cheaper model like Grok 3 Mini provides adequate quality
- Budget-constrained high-volume workloads: Grok 3 Mini Fast offers a strong cost-to-quality ratio in the Grok 3 family
- Tasks that outgrow Grok 3: Grok 4 reports higher scores on hard reasoning benchmarks
Conclusion
Grok 3 Beta is xAI's full-scale Grok 3 reasoning model for tasks where depth of analysis matters. Training on the Colossus supercomputer targets mathematical, scientific, and coding benchmarks. Teams that need the most capable model in the Grok 3 family should start here, trading latency for depth when the task calls for it.
Frequently Asked Questions
What makes Grok 3 Beta different from Grok 2?
Grok 3 Beta is trained on the Colossus supercomputer. It scores higher than Grok 2 on math, science, and coding benchmarks.
What is the context window for Grok 3 Beta?
131.1K tokens. You also get up to 131.1K tokens per response for long completions.
How does Grok 3 Beta compare to Grok 3 Fast?
Grok 3 Beta prioritizes reasoning depth and quality, while Grok 3 Fast optimizes for lower latency at a slight quality tradeoff. Both share the same context window of 131.1K tokens.
What does Grok 3 Beta cost through Vercel AI Gateway?
Current pricing is shown on this page. AI Gateway routes across providers, and rates may vary by provider.
How do I authenticate with Grok 3 Beta through Vercel AI Gateway?
Use your Vercel AI Gateway API key with the model identifier
xai/grok-3. AI Gateway manages provider routing and authentication automatically.Is Grok 3 Beta suitable for agentic applications?
Grok 3 Beta supports tool calling and multi-step reasoning, making it capable for agentic workflows. For latency-sensitive agent loops, consider Grok 3 Fast as an alternative.
Does Vercel AI Gateway support Zero Data Retention for Grok 3 Beta?
Zero Data Retention is not currently available for this model. ZDR on AI Gateway applies to direct gateway requests; BYOK flows aren't covered. See https://vercel.com/docs/ai-gateway/capabilities/zdr for configuration details.