Grok Leads Agent Tests, Gemini Wins on Speed as DeepSeek Raises Prices

Competition among large language models is shifting beyond raw benchmark scores toward a three-way test of intelligence, execution speed and cost. Elon Musk’s xAI, Alphabet’s Google and Hangzhou-based DeepSeek are pitching Grok 4.6, Gemini 3.7 Flash and DeepSeek V4 Pro, respectively, as workhorses for coding and autonomous agents. The comparison matters for businesses because a model that is smarter on paper may still lose if it takes longer, consumes more tokens or costs more to finish a task.

Evaluations published Aug. 12-13 put Grok 4.6 at 70.8% on CursorBench 3.2, while Gemini 3.7 Flash delivered roughly 340 output tokens a second. Google priced Gemini at an introductory $0.75 per million input tokens and $3.75 per million output tokens through Dec. 31, before rates double on Jan. 1, 2027. DeepSeek’s new time-of-day schedule took effect Aug. 17: V4 Pro output costs $1.98 per million tokens off-peak and $3.96 at peak, up from $0.87. The results favor Grok for long-horizon agent work and Gemini for speed and value, while DeepSeek faces criticism over the increase.