Gemini app updates 2.5 Flash with better response formatting

Google today updated the 2.5 Flash model available in the Gemini app across three areas.

All responses now benefit from improved organization and formatting, like the use of headers, lists, and tables. When asking for homework help (especially involving complex subjects), you’ll get “clearer explanations” thanks to improved reasoning capabilities. 

Meanwhile, image understanding now allows you to upload an “even more detailed image or diagram,” and have Gemini explain, organize, or summarize it. The example Google gives is taking pictures of notes and asking for flashcards. 

These updates are already available when you select Gemini 2.5 Flash from the model picker. 

Advertisement – scroll for more content

1/ Clearer explanations for homework

We’ve improved Gemini’s reasoning capabilities for homework questions, so Gemini is much better at walking you through complex subjects with clear, step-by-step guidance.

2/ More scannable outputs

Responses are now better organized with improved formatting. Gemini 2.5 Flash uses elements like headers, lists, and tables to make complex outputs clearer and easier to understand.

3/ Improvements to image understanding

Now you can take a photo of an even more detailed image or diagram, and Gemini can walk you through it. You can even upload your notes and ask Gemini to organize, summarize or create flashcards for you.

At the same, Google today released new preview versions of Gemini 2.5 Flash and 2.5 Flash-Lite for developers. Available in AI Studio and Vertex AI, “today’s releases are not intended to graduate to a new, stable version but will help us shape our future stable releases.”

There are improvements in quality and speed compared to the current stable releases, while there is a “2x improvement in token efficiency and therefore cost.”

gemini-2.5-flash-preview-09-2025:

• Better agentic tool use: We’ve improved how the model uses tools, leading to better performance in more complex, agentic and multi-step applications. This model shows noticeable improvements on key agentic benchmarks, including a 5% gain on SWE-Bench Verified, compared to our last release (48.9% → 54%).

• More efficient: With thinking on, the model is now significantly more cost-efficient—achieving higher quality outputs while using fewer tokens, reducing latency and cost (see charts above).

gemini-2.5-flash-lite-preview-09-2025:

• Better instruction following: The model is significantly better at following complex instructions and system prompts.

• Reduced verbosity: It now produces more concise answers, a key factor in reducing token costs and latency for high-throughput applications (see charts above).

• Stronger multimodal & translation capabilities: This update features more accurate audio transcription, better image understanding, and improved translation quality.