benchmarks
Grok 4 is xAI’s reasoning-focused model for demanding mathematical, scientific and programming tasks, with a documented 256,000-token context window. Newer Grok 4.1 models have shipped, but this profile covers the original Grok 4 rather than Grok 4 Heavy or later generations.
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 | 87.5%Jul 2025 | View |
| Humanity's Last ExamFrontier expert questions across fields | 25.4%Jul 2025 | View |
Weighted mean of 2 verified benchmark results: GPQA Diamond 87.5%, Humanity's Last Exam 25.4%.
Capabilities
- Reasoning across advanced scientific and expert-level questions.
- Solving challenging mathematical problems.
- Generating and editing code across multiple programming languages.
- Processing long documents and other inputs within a 256,000-token context window.
- Calling developer-provided tools through the API.
What improved
- Expands the documented context window from Grok 3’s 131,072 tokens to 256,000 tokens.
- Reports higher GPQA Diamond performance than Grok 3.
- Achieves higher Aider Polyglot performance than Grok 3 Beta.
Known flaws & limitations
- Operates as a reasoning model without a separate non-reasoning mode.
- Does not support the reasoning_effort parameter for adjusting reasoning intensity.
- Still makes substantial errors on difficult expert-level questions; benchmark performance does not establish factual reliability.
Risks, safety & security issues
- Not yet confirmed by public reporting.
Good fit for
- Scientific and mathematical problem-solving with independently checked results.
- Programming assistance and multilingual code editing backed by automated tests.
- Analysis of long technical documents.
- Tool-assisted research workflows with human oversight.
Poor fit for
- High-stakes decisions requiring guaranteed factual accuracy or unsupervised professional judgment.
- Applications requiring an explicit non-reasoning mode.
- Workflows that depend on configuring the reasoning_effort parameter.
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.