▲ Gemini 3.7 Flash
Google has unveiled a new lightweight artificial intelligence (AI) model specialized in coding and agent tasks, but the release of its top-tier model has been delayed once again.
Due to the delayed launch of its Pro model, Google's AI competitiveness appears to be lagging behind latecomer xAI as well.
Google announced on the 13th (local time) that it is introducing the lightweight AI model "Gemini 3.7 Flash" to corporate and paid individual customers.
This new model was revealed just three weeks after the release of "Gemini 3.6 Flash" on July 21.
The company introduced that the new model features significantly improved coding and bug-fixing capabilities, generating highly complete code that can be executed immediately in a single attempt.
In fact, performance benchmarks such as coding for Gemini 3.7 Flash are similar to or even higher than mid-tier models like Anthropic's Claude Sonnet 5 or OpenAI's GPT-5.6 Terra.
Considering that the application programming interface (API) usage price of Gemini 3.7 Flash is two to three times cheaper compared to models like Sonnet and Terra, it concludes that the model offers high cost-efficiency.
In particular, Gemini 3.7 Flash demonstrated unmatched speed in response generation compared to other models.
AI model analysis firm Artificial Analysis (AA) evaluated that the response speed of Gemini 3.7 Flash is 340 tokens per second, making it the second fastest following Gemini 3.5 Flashlight (358 tokens per second).
None of the major models from competitors recorded over 200 tokens per second.
However, due to the limitations of being a lightweight model, Gemini 3.7 Flash ranked lower when compared against all AI models overall.
AA assigned Gemini 3.7 Flash an Artificial Analysis Intelligence Index (AAII) of 56, placing it 9th between GPT-5.6 Terra (57 points) and DeepSeek-V4-Pro-0813 (53 points).
Google has yet to unveil its top-tier model, Gemini 3.5 Pro, which was originally scheduled for release in June, even two months later.
In the meantime, Google has fallen behind not only Anthropic and Open AI, but also Elon Musk's AI firm xAI and Meta, which had been considered latecomers.
Grok 4.6, released by xAI on the 12th, scored 61 on the AAII, tying for 3rd place alongside GPT-5.6 Sol, just 1 point behind Claude 5 Fable (62 points).
Meta's "Muse Spark 1.2" (57 points) and Chinese AI models such as Kimi K3 (60 points) and Qwen 3.8 Max (58 points) also received higher evaluations than Gemini 3.7 Flash.
Consequently, concerns persist that Google is losing its leadership in the AI model market.
(Photo provided by Google, Yonhap News)
※
Copying, redistribution, and unauthorized use in AI training are strictly prohibited.