xAI: Grok 4.3

xAI's Grok 4.3, released on April 30, 2026, is a reasoning model designed for agentic workflows and instruction-following tasks, accepting text and image inputs with text output. While demonstrating exceptional reliability with a 100% success rate across all benchmarks, its performance is characterized by longer response times, ranking in the 18th percentile for speed. Similarly, its pricing tends to be at premium levels, placing it in the 11th percentile for cost-effectiveness. Despite these speed and cost considerations, Grok 4.3 exhibits remarkable accuracy across several critical domains. It achieved perfect scores in Hallucinations (100%), General Knowledge (100%), Email Classification (100%), and Ethics (100%), often being the most accurate model at its respective price point and speed for these categories. Its instruction-following capabilities are strong at 78% accuracy, and it performs very well in Coding (96%) and Mathematics (95%), indicating robust problem-solving and domain-specific knowledge. Reasoning tasks also show high accuracy at 96%. The model's key strengths lie in its high accuracy across a broad spectrum of knowledge and reasoning tasks, coupled with its perfect reliability. Its primary weakness is its slower processing speed and premium pricing, which may impact its suitability for latency-sensitive or budget-constrained applications.