Quick Take

Sapiens AI released the Agnes 2.5 Pro Alpha on July 24, 2026, positioning it as a specialized, budget-friendly model. In contrast, Google’s Gemini 3.5 Flash (medium), released on May 19, 2026, serves as a high-performance, high-speed model backed by Google’s extensive infrastructure. While Agnes excels in cost efficiency, Gemini leads in raw intelligence and throughput.

Benchmark Read

Gemini 3.5 Flash (medium) consistently outperforms Agnes 2.5 Pro Alpha across available metrics. Gemini records an Intelligence index of 45.4 compared to Agnes’s 38.8. In shared benchmarks, Gemini leads with a GPQA score of 0.921 (vs. 0.876), an HLE score of 0.399 (vs. 0.319), a SciCode score of 0.53 (vs. 0.422), and an LCR score of 0.71 (vs. 0.637). Agnes 2.5 Pro Alpha reports a Coding index of 58.8, whereas this metric is unknown for Gemini.

Cost and Speed

There is a stark contrast in pricing. Agnes 2.5 Pro Alpha is significantly more affordable, with a blended cost of $0.56/1M tokens (Input: $0.45, Output: $0.90). Gemini 3.5 Flash (medium) is priced at a blended $3.38/1M tokens (Input: $1.50, Output: $9.00).

Regarding performance, Gemini 3.5 Flash (medium) delivers an output speed of 246.743 tok/s, nearly double the 133.925 tok/s of Agnes. However, Agnes 2.5 Pro Alpha offers a much faster time to first token at 1.929s, compared to Gemini’s 12.828s.

Best Fit

Agnes 2.5 Pro Alpha is best suited for cost-sensitive projects and coding tasks where budget optimization is the primary driver. Gemini 3.5 Flash (medium) is the ideal candidate for high-demand, agentic workflows requiring rapid output and high intelligence, particularly where the higher cost is offset by performance gains.