benchmarks
Grok 3 Beta is xAI’s text-based API model for conversational assistance, coding, mathematics and scientific questions. Grok 4 launched in July 2025 as a newer generation; the specifications here remain specific to Grok 3 Beta.
Benchmark results
Every figure links to where it was published. Compare like with like — harnesses and prompt setups differ between vendors.
| Benchmark | Score | Source |
|---|---|---|
| GPQA DiamondPhD-level science questions | 75.8%26 May 2025 | View |
| OTIS Mock AIME 2024-25Competition mathematics | 55.6%10 Apr 2025 | View |
| FrontierMath | 3.8%10 Apr 2025 | View |
| MATH Level 5 | 88.7%10 Apr 2025 | View |
| Aider PolyglotCode editing across six programming languages | 53.3%10 Apr 2025 | View |
Weighted mean of 5 verified benchmark results: GPQA Diamond 75.8%, OTIS Mock AIME 2024-25 55.6%, FrontierMath 3.8%, MATH Level 5 88.7%, Aider Polyglot 53.3%.
Capabilities
- Text-based conversation, drafting and question answering.
- Code generation, explanation and editing.
- Mathematical problem solving.
- Scientific question answering.
- Function calling for integration with developer-provided tools.
What improved
- xAI reports stronger mathematical benchmark performance than Grok 2.
- xAI reports stronger scientific question-answering performance than Grok 2.
- xAI reports stronger coding benchmark performance than Grok 2.
Known flaws & limitations
- The API model accepts text rather than native image, audio or video inputs.
- Its documented knowledge cutoff is November 2024; newer information requires external sources.
- The Grok 3 API model does not expose the reasoning-effort control available in Grok 3 Mini.
Risks, safety & security issues
- Not yet confirmed by public reporting.
Good fit for
- Text-based assistants connected to developer-provided tools.
- Code drafting, explanation and editing with testing and review.
- Mathematical and scientific assistance with independent verification.
- Summarization and analysis of supplied text documents.
Poor fit for
- Applications requiring native image, audio or video understanding.
- Current-events answers without retrieval or other up-to-date sources.
- Workflows requiring adjustable reasoning effort.
References
Profile last reviewed 10/27/2026 and re-checked automatically every few hours against the vendor's published documentation, papers and public leaderboards. Scores come from the cited benchmark results above — confirm limits and pricing against the provider's own documentation before building on them.