Grok 4.3

Grok 4.3 debuts with strong multimodal performance across benchmarks

Grok 4.3 enters the benchmark landscape as a competitive multimodal model with solid performance across core evaluation suites. The model achieves 87.8% on MMLU, placing it in the upper tier of current language models, alongside 88.3% on MMLU Pro and 78.6% on GPQA Diamond, demonstrating strong reasoning capabilities. Mathematical performance is robust with 91.6% on MATH-500 and 90.1% on GSM8K, while coding abilities show well with 89.7% on HumanEval and 84.5% on LiveCodeBench. Vision capabilities debut at 78.9% on MMMU, indicating competent multimodal understanding. Context handling extends to 128K tokens with function calling support, making it suitable for complex applications. The model shows balanced performance without obvious weaknesses in core domains. As a first-generation baseline, Grok 4.3 establishes itself as a capable general-purpose model with particular strength in mathematical reasoning and coding tasks. Users should find it well-suited for applications requiring both text and vision processing, though performance in specialized domains will need evaluation in real-world scenarios.