As we all know, the development of artificial intelligence is so fast now that major technology companies are holding back their big moves, just hoping to take the lead in the field of models. On March 4, 2026, DeepMind, a subsidiary of Google, officially launched the Gemini3.1 Flash Lite preview version, which immediately caused a stir in the tech industry.
What is Gemini3.1 Flash Lite Preview?
Gemini3.1 Flash Lite Preview is a member of the Gemini3 series. You can imagine it as the "speed expert" in this series, and the cost-effectiveness is quite high. It is an upgraded version of Gemini 2.5 Flash Lite, just like upgrading a phone from the first generation to the second generation, with improvements in many aspects.
Performance improvement: It's like hanging up
The speed is still impressive
Let's talk about speed first. This model inherits the advantages of its predecessor, with the ability to output over 360 tokens per second and an average response time of only 5.1 seconds. What concept? Just like when you say a sentence, it can respond to you in the blink of an eye. It is particularly fast when processing large amounts of text information, such as online customer service chat and real-time translation, making it particularly suitable for use.
Getting smarter
At the level of intelligence, its progress is not insignificant. An organization called Artificial Analysis Intelligence Index Monitoring gave it a score that was 12 points higher than its predecessor, reaching 34 points.
In the Arena.ai ranking, its Elo score is 1432, which indicates that the content it generates is of higher quality and more in line with our human needs, just like a student who used to have average grades suddenly becoming a top student.
Super powerful core ability
It performs particularly well in key abilities such as multimodality and scientific reasoning.
In the GPQA Diamond test, it achieved a high score of 86.9%, which is like getting close to a perfect score on an exam, indicating its strong ability to handle complex problems and deep reasoning.
In the MMMU Pro benchmark test, the accuracy reached 76.8%, even surpassing heavy models such as Claude Opus 4.6 and Kimi K2.5.
In the future, scientific research and financial analysis, which require high data accuracy and reasoning, can be of great help.
Can adapt to various scenarios
This model also has a great feature, which is that developers can adjust the depth of its "thinking" according to their own needs. Whether it's simple automated translation or complex UI construction, it can easily handle it, like a universal tool that can do anything.
Price adjustment: The price has risen a bit sharply
However, as the performance improved, the price also increased significantly. Previously, for every million input tokens, it was only 0.25 US dollars (based on the original logic, it is speculated that the previous generation's price was incorrect, and the adjusted input price will be used uniformly, but further verification may be needed in reality). Now it is still at this price; But the output price is quite different, it has suddenly increased from $0.40 to $1.50, almost tripling.
It's like when you go shopping and the item gets better, the price also increases significantly. This is because the performance of the model has improved, and the costs of research and development, training, and operation have also increased. Google has no choice but to raise prices to maintain business.
Marketing: The gameplay has changed
In the past, the lightweight model market was like a vegetable market, where everyone was competing for lower prices to attract customers. But now it's different, with the Gemini3.1 Flash Lite preview version starting to be tested on Google AI Studio and Vertex AI, the market trend has changed. Google uses this high-performance model to tell everyone that in the field of artificial intelligence, low price alone is not enough, good model performance and high quality are the key. This move can not only attract users who have high requirements for the model, but also make the entire industry pay more attention to technological innovation and product quality, promoting the market to develop in a better direction.
Overall, the performance improvement of the Gemini3.1 Flash Lite preview version launched by Google is quite significant. Although the price has increased, its capabilities have also increased significantly. Let's wait and see how big a wave it can create in the field of artificial intelligence in the future!
