Google says that its improved Gemini 2.0 Flash AI model is now generally available to everyone. The API for Google's best models comes at a lower price tag than DeepSeek-R1, the Chinese chatbot that stirred Silicon Valley.
The new Gemini 2.0 Flash was first introduced in December 2024. Now, this model has been updated with a “thinking” feature and is available to everyone. Users can interact with it via the Gemini app or the API in Google AI Studio and Vertex AI.
“We’re also releasing an experimental version of Gemini 2.0 Pro, our best model yet for coding performance and complex prompts,” Google announced.
The Pro version is available for subscribers (Gemini Advanced users) and in Google AI Studio and Vertex AI.
Google’s chatbots score high across multiple benchmarks. However, their performance seems lower than DeepSeek's self-estimates. The Chinese startup released the open-source R1 model a few weeks ago.
For example, the Gemini 2.0 Flash scores 77.6% on the MMLU-Pro test, which evaluates knowledge across multiple subjects. DeepSeek-R1 boasts an 84% score.
In GPQA Diamond, a very challenging dataset with questions from biology, physics, and chemistry, Gemini 2.0 Flash scores 60%, the Pro version reaches almost 65%, while the DeepSeek-R1 achieved 71.5%.
However, Google doesn’t provide benchmark results for the “thinking” flavors of its models, which are now available and may perform better. In Chatbot Arena, 2.0 Flash Thinking Experimental is currently ranked the world’s best.
Google says its Germini 2.0 Pro “has the strongest coding performance and ability to handle complex prompts, with better understanding and reasoning of world knowledge, than any model we’ve released so far.” Until further testing is available, the R1 appears to perform better in coding, according to BigCodeBench.
Gemini outshines R1 and most of the other models with a gigantic context window of two million tokens for the Pro Experimental version and one million for the other Flash models. The large context window enables “to comprehensively analyze and understand vast amounts of information, as well as the ability to call tools like Google Search and code execution.”
Google now also offers better Gemini 2.0 Flash API pricing for developers. It is priced at $0.10 for a million input tokens (DeepSeek-R1 - $0.14-$0.55/1M input tokens) and $0.4 for a million output tokens (DeepSeek-R1 – $2.19/1M output tokens). Google hasn’t listed the pricing for the Pro model. Google models also support image and audio inputs.
Google also introduced a more cost-efficient alternative – the Gemini 2.0 Flash-lite model, which outperforms the previous 1.5 Flash. This model costs $0.075 and $0.3 per million input and output tokens, respectively.
“Like 2.0 Flash, it has a one million token context window and multimodal input. For example, it can generate a relevant one-line caption for around 40,000 unique photos, costing less than a dollar in Google AI Studio’s paid tier,” Google said.
Your email address will not be published. Required fields are markedmarked