Report says Alphabet’s latest Gemini model nearly matches Anthropic on a key AI benchmark at about one-third the cost

Alphabet is closing in on Anthropic in an industry benchmark for demanding AI performance, according to a market report published on Nov. 4, 2026. The article said Google’s latest Gemini model delivered almost the same results on the test as Anthropic’s model while operating at roughly one-third the cost, a claim that, if accurate, would underscore an intensifying focus on performance per dollar rather than raw capability alone.

The report framed the comparison as part of a “new benchmark” that pushes AI systems in difficult tasks. It also characterized the gap between the two companies as narrowing, stating that Alphabet’s latest model nearly matches Anthropic’s output under the benchmark’s scoring criteria.

While the article emphasized cost efficiency, it did not provide enough detail in the information available here to verify how “one-third the price” was calculated, such as whether the figure refers to inference cost, training cost, or another cost basis. It also did not specify the exact benchmark suite, the model versions compared, or whether costs were measured under identical hardware and latency assumptions.

For Alphabet, Gemini is the company’s family of AI models used across products and services, from developer tools to assistant and productivity experiences. In practice, improvements that keep quality high while reducing compute demand can matter because AI workloads are expensive at scale, and even small cost reductions can translate into meaningful changes in operating expense and product margins.