Grok 4.20 Beta Reasoning
Grok 4.20 Beta Reasoning is xAI's reasoning model in the Grok 4.20 beta generation, supporting extended thinking for complex problem-solving with low hallucination rates and strict prompt adherence.
import { streamText } from 'ai'
const result = streamText({ model: 'xai/grok-4.20-reasoning-beta', prompt: 'Why is the sky blue?'})Playground
Try out Grok 4.20 Beta Reasoning 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.
Grok 4.20 Beta Reasoning
Ask Grok 4.20 Beta Reasoning anything to try it out.
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.
| Provider |
|---|
xAI | 2M | 0.4s | 237tps | $1.25/M | $2.50/M | Read: $0.2/M Write: — | $5/K + input costs | — | +4 | 03/11/2026 |
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 |
|---|
| 256K | 0.4s | 131tps | $1.00/M | $2.00/M | Read: $0.2/M Write: — | $5/K + input costs | — | +3 | | 05/20/2026 | |||
| 1M | 1.0s | 151tps | $1.25/M | $2.50/M | Read: $0.2/M Write: — | $5/K + input costs | — | +4 |  | 04/30/2026 | |||
| 2M | 0.4s | 152tps | $1.25/M | $2.50/M | Read: $0.2/M Write: — | $5/K + input costs | — | +3 | | 03/10/2026 | |||
| 2M | 0.4s | 140tps | $1.25/M | $2.50/M | Read: $0.2/M Write: — | $5/K + input costs | — | +4 | | 03/10/2026 | |||
| 1M | 0.3s | 122tps | $0.20/M | $0.50/M | Read:$0.05/M Write:— | — | — | +2 | | 11/19/2025 | |||
| 1M | 1.1s | 121tps | $0.20/M | $0.50/M | Read:$0.05/M Write:— | — | — | +3 | | 11/19/2025 |
About Grok 4.20 Beta Reasoning
Grok 4.20 Beta Reasoning was released March 11, 2026 as the beta-tagged variant of xAI's Grok 4.20 reasoning model. It generates visible chain-of-thought reasoning before producing its final answer, letting it work through multi-step problems, mathematical proofs, and complex code analysis more reliably than direct-response models.
The reasoning architecture adds latency per request but improves accuracy on tasks that benefit from deliberation: debugging intricate code paths, synthesizing information across long documents, and solving problems that require intermediate steps. Combined with the Grok 4.20 generation's low hallucination rates, this makes Grok 4.20 Beta Reasoning suitable for high-stakes analysis where correctness matters more than speed.
As a beta release, Grok 4.20 Beta Reasoning may receive weight or behavior updates. For stable reasoning workloads, use the non-beta grok-4.20-reasoning.
What To Consider When Choosing a Provider
- Configuration: Grok 4.20 Beta Reasoning is in beta. Expect potential changes before general availability.
- Configuration: Extended thinking adds tokens and latency to each response. For tasks that don't benefit from structured reasoning, the non-reasoning variant is faster and cheaper.
- 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 4.20 Beta Reasoning
Best For
- Complex problem solving: Requiring multi-step reasoning and extended thinking
- Mathematical and scientific analysis: Structured chain-of-thought improves accuracy
- Code debugging and review: That benefits from step-by-step logical analysis
- High-stakes decisions: Low hallucination rates and traceable reasoning are critical
- Research and analysis tasks: Operating over large document collections within the context of 2M tokens
Consider Alternatives When
- Speed-focused tasks: The Grok 4.20 Non-Reasoning variant delivers faster responses at lower cost
- Multi-agent orchestration: The Grok 4.20 Multi-Agent variant is purpose-built for agent collaboration
- Stable production deployments: Beta models introduce risk. Use Grok 4.1 Fast Reasoning or the full Grok 4 instead
- Simple direct-response tasks: Reasoning overhead adds cost without improving quality
Conclusion
Grok 4.20 Beta Reasoning adds extended thinking to the Grok 4.20 beta generation. Use it for complex analysis, multi-step reasoning, and code debugging where deliberation depth improves accuracy. For production stability, prefer grok-4.20-reasoning.