Google Gemini 2.5 (Flash/Pro/Nano Banana)

[limmk]

 

[limmk]

 

[limmk]

 

[limmk]

 

[limmk]

 

[limmk]

As promised last week, Google is rolling out access to Gemini 2.0 Flash Experimental in the Android app.

Initially, 2.0 Flash Experimental was only available on gemini.google.com, which has long featured a model picker in the top-left corner.

Version 15.50 of the Google app is rolling out to the beta channel this morning with a model switcher that goes beyond just enabling Gemini Advanced from settings.

On the homepage, the model information introduced last week is now tappable and lets you pick:

• 1.5 Pro: Tackle complex tasks
• 1.5 Flash: Get everyday help
• 2.0 Flash Experimental: Preview our latest experimental model

Google warns that 2.0 Flash Experimental “Might not work as expected,” while file upload is not currently available. Gemini 2.0 Flash “outperforms 1.5 Pro on key benchmarks” at twice the speed. For Gemini app users, Google says the new model results in an “even more helpful Gemini assistant.”

While we’re excited for you to try it, remember it’s an early preview and might not work as expected. Additionally, some Gemini features won’t be compatible with this model in its experimental state.

With this change, Google has removed the old “Advanced” toggle from Gemini settings.

Google app 15.50 to access Gemini 2.0 Flash should hit the stable channel in the next week, or you can switch to the beta here. The experience has historically been pretty stable

We’re not seeing the model picker in the iOS app just yet. Meanwhile, Gemini Advanced’s Deep Research feature is coming to mobile in early 2025.

 

[limmk]

Following Gemini 2.0 Flash Experimental in the past week, Google is now letting Gemini Advanced customers access a new 2.0 Experimental Advanced model.

Appearing as “2.0 Experimental Advanced” in the model picker on desktop and mobile web (not app), you’re specifically using “Gemini-Exp-1206,” which Google says offers “significantly improved performance on complex tasks such as coding, math, reasoning and instruction following.”

Handling “complex tasks with greater ease,” it can be used for “tackling complex coding challenges, solving mathematical problems for school or personal projects, or providing detailed, multi-step instructions to craft a tailored business plan.”

As an “early preview,” Google warns that it “might not work as expected,” while this model does not have access to real-time information and “won’t be compatible with some Gemini features in its experimental state.”

2.0 Experimental Advanced. Lacks access to real-time info and some Gemini features.

Gemini-Exp-1206 is currently at the top of the Chatbot Arena LLM Leaderboard.

To date, Gemini Advanced has given subscribers access to 1.0 Ultra and 1.5 Pro models. As such, it’s possible that Gemini-Exp-1206 is 2.0 Pro or higher. Google said to expect more Gemini 2.0 model sizes in January when 2.0 Flash hits general availability for developers.

Gemini Advanced requires the $19.99 per month Google One AI Premium subscription.

 

[limmk]

Google today released an experimental “Gemini 2.0 Flash Thinking” model that “explicitly shows its thoughts” to solve complex problems.

As the name suggests, it is built on “2.0 Flash’s speed and performance.” Google says it is “trained to think out loud,” thus “leading to stronger reasoning performance.”

Competing with OpenAI’s o1, Google shared several demos across physics and probability:





Gemini 2.0 Flash Thinking is available in Google AI Studio (direct link) and Vertex AI today. You can click “Expand to view model thoughts” and see the reasoning occur in real-time before it provides the final answer. This is “just the first step in [Google’s] reasoning journey.”

It has debuted at “#1 across ALL categories” on the Chatbot Arena LLM Leaderboard. Just yesterday, Google launched made 2.0 Experimental Advanced available in the Gemini app, with Gemini-Exp-1206 also at the top of the leaderboard.

---

The leap from Gemini-2.0-Flash:

• Overall: #3 → #1
• Overall (Style Control): #4 → #1
• Math: #2 → #1
• Creative Writing: #2 → #1
• Hard Prompts: #1 → #1 (+14 pts)
• Vision: #1 → #1 (+16 pts)

---

It remains to be seen how this will ultimately launch for end users. These reasoning capabilities will presumably be integrated into the main model down the road, with Google’s framing as being part of the Gemini 2.0 family a good indicator of that. At the moment, we already have a task-specific model with “1.5 Pro with Deep Research.”

 

[limmk]

After hitting the Google app beta channel on Monday, access to 2.0 Flash Experimental in the Gemini app on Android is now rolling out to stable users.

The Gemini app is powered by the Google (Search) app on Android. Version 15.50 is now rolling out to the stable channel via the Play Store.

If you don’t immediately see the model switcher/dropdown at the top of the Gemini homepage after updating, Force stop the Google/Gemini app from the App info page.

2.0 Flash access is available to both free Gemini and Advanced users on Android. The model picker shows an “Upgrade” promo for the subscription. Your choices are:

• 1.5 Pro: Tackle complex tasks
• 1.5 Flash: Get everyday help
• 2.0 Flash Experimental: Preview our latest experimental model

The Gemini app notes that 2.0 Flash Experimental “Might not work as expected,” while file upload is not currently available for any tier. Gemini 2.0 Flash “outperforms 1.5 Pro on key benchmarks” at twice the speed. In the Gemini app, this new model results in an “even more helpful Gemini assistant.” 2.0 Flash will see general availability for developers in January, and more model sizes are coming.

While we’re excited for you to try it, remember it’s an early preview and might not work as expected. Additionally, some Gemini features won’t be compatible with this model in its experimental state.

Google updated the Gemini app on iPhone with 2.0 Flash access yesterday. Mobile users should also be on the lookout for Deep Research access early next year.

 

[limmk]

Earlier this month, Google started testing Gemini in Google Assistant on the Nest Mini and Audio, and here are the new voices.

There are seven Gemini in Google Assistant voices to choose from. They are named after plants. Like Gemini Live’s 10 options, they are “more natural sounding voices.”

• Aloe: Calm • Mid-range voice
• Oxalis: Bright • Mid-range voice
• Fern: Bright • Higher voice
• Verbena: Calm • Deeper voice
• Ivy: British accent • Mid-range voice
• Jade: Engaging • Mid-range voice
• Eucalyptus: Australian Accent • Higher voice

To see if you have this, enroll in the Google Home Public Preview program. Afterwards, you might have an “Experimental AI features” toggle on that page. This setting is not yet widely rolled out even if you have other Public Preview features and is also responsible for AI camera search.

Once rolled out, go to the Google Home Settings tab > Google Assistant > Assistant voice & sounds.

When you’re using Gemini, there will be a “chime before Google Assistant responds with an AI-powered answer.” This large language model can “respond to a wider range of questions and provide more in-depth, AI-powered answers on general knowledge topics.” You can also ask follow-up questions, with the ability to interrupt Assistant’s response at any time.

 

[limmk]

TL;DR

• Google could be about to give users new settings to control Gemini responses.
• This “content filter” would presumably offer control over the extent to which Gemini censors itself.
• The Gemini API already has settings for “harm categories” accessible to developers.

Artificial intelligence has been working its way into practically everything this past year, and when it comes to AI and Google, Gemini is the name of the game. The company’s been working hard at improving its models, and earlier this month announced Gemini 2.0. Because of the sort of power AI offers, companies like Google dedicate a lot of time to thinking about the sort of protections their AI systems need to have in place. Right now, we’re taking a look at a new way that Gemini may soon give users more control over the sort of content it generates.

Today we’re checking out version 15.51.24.sa.arm64 beta of the Google app for Android. While this isn’t yet publicly facing and live just yet, we were able to activate a new option under Gemini settings that appears to be preparing for the ability to filter Gemini output by content.

At the moment, selecting this option doesn’t do anything useful, and only takes us to a URL on Google’s Gemini site that’s not yet up. While that makes it a bit difficult to say with much certainty how this setting is likely to work, we can make a few inferences based on what we already know about Gemini.

Google already has some formal policies in place about what it’s OK to do with Gemini, and extending from that, also has some goals for the sort of content it tries to steer Gemini towards avoiding. Developers, in particular, have had access to some reasonably fine-grained Gemini safety settings, giving them control over a number of “harm categories” including hate speech, harassment, and sexually explicit content.

Knowing that all that’s currently available, it might be possible that Google could be extending some level of that same control to end users of Gemini in Android and on the web. That said, Google could go in any number of directions when it comes to its implementation here, from the sort of broad, developer-level filters the API has now, to ones that are maybe even more restrictive, all in the interest of preventing Gemini from sharing anything particularly repugnant.

Right now, it’s hard to say when this new option might go live, but we’ll be keeping an eye out for any further motion.

 

[limmk]

 

[limmk]

After introducing in mid-December, Gemini Advanced subscribers can now access 2.0 Experimental Advanced in the Gemini app on Android and iOS.

While Gemini 2.0 Flash Experimental has been available for all users since last month, Google has just introduced 2.0 Experimental Advanced to the Gemini app. It was previously only available on the desktop/mobile website.

Gemini-Exp-1206 offers “significantly improved performance on complex tasks such as coding, math, reasoning and instruction following.” However, it “lacks access to real-time info and some Gemini features,” like file upload.”

The model, which remains at the top of the Chatbot Arena LLM Leaderboard, can be used for “tackling complex coding challenges, solving mathematical problems for school or personal projects, or providing detailed, multi-step instructions to craft a tailored business plan.”

Tap the model switcher at the top of your screen. 2.0 Experimental Advanced joins 1.5 Flash, 1.5 Pro, and 2.0 Flash Experimental. Force stop the Gemini/Google app on Android if you’re not seeing it yet, while Google released an update to the Gemini app for iOS today following last month’s bigger upgrade.

Gemini Advanced requires a Google One AI Premium subscription ($19.99 per month).

It remains to be seen what model Gemini-Exp-1206 will launch as. It could be Gemini 2.0 Pro or some other model above 2.0 Flash.

 

[limmk]

Google has been slowly rolling out a redesign of Gemini overlay on Android and more beta users are now seeing it as of Monday morning.

The Gemini overlay that opens when you say “Hey Google,” long-press the power button, or swipe up from the bottom corners has shrunk a great deal with this redesign. On the Pixel Launcher, it perfectly covers the Google Search bar. It’s now a text field that just has a ‘plus’ menu, microphone, and Gemini Live shortcut.

Old vs. new

There’s a prominent blue/purple glow around the overlay and circular voice button. You tap “Ask Gemini” for text input, while the camera shortcut has been moved into the plus. Google has removed the “Good morning, afternoon, or evening” greeting, as well as the shortcut to open the full app in the top-right corner and the drag handle to enter split-screen multitasking.

This overlay can expand to a maximum of four full lines of text for particular long queries.

There are also tweaks to the Gemini app’s homepage. The Ask Gemini field moves the camera shortcut into the ‘plus’ menu: Camera, Gallery, Files, and Drive. (The last two options are only available for Gemini Advanced.)

Additionally, the Gemini Live shortcut has been moved into the pill-shaped bar, which is now thinner and removes the separator line above it. Google has also gotten rid of the more elaborate text input animation for a simple circle around the microphone when it’s active.

As of Monday, we’re seeing this overlay redesign after force stopping the Gemini/Google app on quite a few Pixel devices — but not Samsung or OnePlus — enrolled in the Google app beta (version 16.1.43). It’s not yet consistently appearing on phones running the stable channel version of the app. This is showing up on free Google Accounts, Gemini Advanced, and Workspace accounts.

 

[limmk]

Back in December, Gemini 2.0 Flash Thinking was announced as Google’s first reasoning model and an updated experimental version is now available to test.

Built on 2.0 Flash, which was announced earlier that month, this model “explicitly shows its thoughts” (like in AI Studio) for better reasoning performance, with the ability to solve more complex problems. It joins gemini-2.0-flash-exp and gemini-exp-1206.

Google today released Gemini 2.0 Flash Thinking Experimental 01-21 (the hyphen between the month and day is new for Google’s model naming scheme) with:

• 1 million token context window (from 32k): This is useful for those that want to “plug in a codebase or query a set of papers with more complex reasoning”
• Native code execution support: For improved tool use
• Longer output token generation
• Less frequent model contradictions (“reduced likelihood of thought-answer contradictions”)

Compared to Exp 1219, it has “better performance on math, science, and multimodal reasoning benchmarks,” including 73.3% on the AIME2024 (Math) and 74.2% on GPQA Diamond (Science) benchmarks.

DeepMind CEO Demis Hassabis said this “represents super fast progress from our first release just this past Dec.” More broadly:

We’ve been pioneering these types of planning systems for over a decade, starting with programs like AlphaGo, and it is exciting to see the powerful combination of these ideas with the most capable foundation models.

Gemini 2.0 Flash Thinking Experimental is free to test in Google AI Studio and via the API. Looking ahead, Sundar Pichai says the “progress on scaling thinking is incredible and will continue to iterate,” with “more to come.”

Meanwhile, Google earlier this month updated the mobile UI for AI Studio:

 

[limmk]

 

[limmk]

 

[limmk]

 

[limmk]

After first coming to the Public Preview program, the Google Home Gemini Extension is now rolling out to all users. Google tells us that it will be fully available over the coming weeks.

Accessible on Android and iOS (Gemini app > Extensions menu > @google Home), you can control your smart home in a more conversational manner. Replacing the Google Assistant fallback, this includes:

• Lights & power, like lights, outlets, and switches
• Climate control, like air conditioning units, thermostats, heaters, and fans
• Window coverings, like curtains, blinds, and shutters
• Media devices, like TVs and speakers
• Other smart devices, like washers, coffee makers, and vacuums

Example commands are listed below.

Meanwhile, the Google Home Gemini Extension has been updated with the ability to control non-sensitive devices, like smart lights, from your phone’s lock screen (without unlocking).

The Gemini app now lets you adjust the volume, pause, and resume media playing on TVs, speakers, displays, and other connected entertainment devices.

Thermostat tile controls have been updated “with warm tones when the heat is on and cool tones when you’re running the A/C,” thus matching the Google Home app.

Meanwhile, Gemini will automatically open the Google Home app for lock features, camera, and other supported actions

Finally, Google is also touting “manage multiple commands or complex queries and ask questions about the status of your devices.”

• “Turn on all the living room lights except the armchair light.”
• “Actually, turn the armchair light on too, but dim the kitchen lamp.”
• “Is the back porch light still on?”

---

Lights​

• Turn on/off [light name].
• Turn on/off all of the lights.
• Turn my [room name] lights on/off.
• Dim the [light name].
• Dim the [room name] lights.
• Brighten the [light name].
• Set [light name] to 50%.
• Turn [light name] green.

Switches or outlets​

• Turn on/off [outlet name].
• Turn on/off [switch name].

Thermostats​

Turn on or off

• Turn on heating/cooling mode.
• Turn on heat-cool mode.
• Turn off thermostat.

Set or adjust the temperature

• Set the heat to [temperature].
• Set heat-cool to [temperature].
• Set the air conditioning to [temperature].
• Set the [room name] thermostat to [temperature].
• Make it warmer/cooler.
• Raise/lower the temp.
• Raise/lower the temp by 2 degrees.

Switch heating or cooling modes

• Turn on heating/cooling.
• Set thermostat to cooling/heating.
• Turn thermostat to heat-cool mode.

Fans, heaters & A/C units​

• Turn on/off [fan, heater, A/C, device name].
• Increase the temperature on my heater
• Increase/decrease the fan speed

Control window coverings​

• Open/close [curtain name]
• Open/close [blinds name]
• Open/close [shutters name]

Control media devices​

• Turn on/off [TV name]
• Turn volume up/down on [TV name, speaker name]
• Pause/resume [TV name, speaker name]

Control other smart home devices​

Other devices, like a vacuum, washer, coffee maker & more

• Start [device name]
• Stop [device name]
• Vacuum the [room]
• Turn on/off [device name]

 

[limmk]

Google is widely rolling out a redesign of the Gemini overlay and homepage on Android.

The Gemini overlay appears at the bottom of your screen when you say the “Hey Google” hotword, long-press the power button, or swipe up from the bottom corners.

Old vs. new

This redesign makes it much smaller, and simplifies the panel down to a ‘plus’ menu (for Camera and Gallery), the “Ask Gemini” field that you can tap for text input, microphone, and Gemini Live shortcut. Google has removed the split-screen multitasking handle and the ability to open the full Gemini app. The Google Assistant-esque “Good morning, afternoon, or evening” greeting is also gone.

As what you’re saying is transcribed, the text field will get taller, while the “Ask about screen,” “Ask about PDF,” and “Ask about YouTube” suggestion chips that appear above the overlay are now left-aligned instead of being centered.

The blue/purple glow around the perimeter is quite vibrant at this size, while the ring around the microphone replaces the old Material You shape for voice input.