benchmarks
Grok 3 Mini Beta is xAI’s text-only reasoning model designed for mathematics, coding and logic-focused tasks, with configurable reasoning effort and separate reasoning output. Later Grok generations have shipped; this profile remains specific to the original Grok 3 Mini beta API model.
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 | 76.3%10 Apr 2025 | View |
| OTIS Mock AIME 2024-25Competition mathematics | 77.8%10 Apr 2025 | View |
| FrontierMath | 5.9%10 Apr 2025 | View |
| MATH Level 5 | 90.9%10 Apr 2025 | View |
| LMArena TextHuman preference Elo from blind chat votes (not a percentage) | 1364 Elo22 Oct 2026 | View |
| Aider PolyglotCode editing across six programming languages | 49.3%10 Apr 2025 | View |
Weighted mean of 5 verified benchmark results: GPQA Diamond 76.3%, OTIS Mock AIME 2024-25 77.8%, FrontierMath 5.9%, MATH Level 5 90.9%, Aider Polyglot 49.3%.
Capabilities
- Reasoning-oriented responses for mathematical and logical problems.
- Code generation and assistance with programming problems.
- Low and high reasoning-effort settings for balancing computational effort against response speed.
- Separate reasoning_content and final-answer fields in API responses.
What improved
- Adds explicit reasoning before answering compared with the non-reasoning Grok 2 generation.
- Introduces configurable reasoning effort for adjusting problem-solving depth.
- Makes reasoning content available separately from the final answer through the API.
Known flaws & limitations
- Accepts text rather than native image, audio or video input.
- Higher reasoning effort can increase latency and reasoning-token consumption.
- Positioned for logic-based tasks rather than work requiring extensive domain-specific knowledge.
Risks, safety & security issues
- Not yet confirmed by public reporting.
Good fit for
- Mathematical problem-solving and quantitative reasoning.
- Programming assistance and algorithmic problems.
- Logic-focused workflows with adjustable reasoning effort.
- Cost-sensitive text applications that benefit from reasoning.
Poor fit for
- Applications requiring native image, audio or video understanding.
- Specialist knowledge tasks that depend on extensive factual recall.
- Strict low-latency workflows using high 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.