Google's Tensor G5: A Leap Forward in AI-Powered Mobile Computing

Pixel 10, Pixel 10 Pro, and Pixel 10 Pro XL debut as Google's most ambitious lineup yet

Google's got more to prove than ever this year, with all eyes on the Tensor G5 and its performance. Unsurprisingly, this chip is heavily focused on AI capabilities. It can run on-device models with the help of an upgraded TPU, which Google says is 60% more powerful than before. That new production node doesn't just help the Tensor G5 run on average 34% faster but also aids in keeping power consumption down. And in case you haven't picked up on it from those incredible camera tricks we were just touching on, this chip's got an upgraded image processor on board.

Google did more than just switch to TSMC for Tensor G5 -- This is what's new

The Pixel 10 series is powered by the Tensor G5, Google's biggest upgrade to its custom silicon. Google touts deeper customization, starting with how it's manufactured by TSMC on the latest 3nm process node. This results in a CPU that is 34% faster on average, with "significant gains" in single and multi-threaded performance. Google says to expect improvements in responsiveness when browsing the web, launching apps, OS rendering, AI experiences, and other workloads. There's one big performance core, five mid-performance cores, and two efficiency cores. Google has also upgraded hardware and software thermal controls to let the chip operate at higher frequencies without throttling. Tensor G5 has a high-speed memory interface thanks to LPDDR5X (higher memory bandwidth) and UFS 4.0 (faster flash storage). The TPU (Tensor Processing Unit) is up to 60% more powerful, while Tensor G5 runs the "newest" Gemini Nano model from DeepMind. This translates to Gemini Nano running "2.6x faster and 2x more efficiently for use cases like Pixel Screenshots and Recorder" compared to Tensor G4. There's also a 32K token context window on G5 (which can be a month's worth of emails or a hundred screenshots) compared to 12K last year. The Tensor team touts collaboration with DeepMind on improving the quality of on-device models. G5 makes use of the Matformer Model Architecture to more efficiently run models, while Per Layer Embedding improves model response quality given the constraints of RAM on mobile devices. Tensor G5 is responsible for over 20 on-device AI features at launch, like Magic Cue, Call Notes with actions, the new Journal app, Scam Detection, and Gboard Smart Edit. In the case of Voice Translate, the translation of what the speaker is saying involves a generative model running in parallel with a classical audio ML model. One-shot voice preservation means Google is able to reconstruct the speaker's vocal characteristics from just a few seconds of live audio. On the privacy front, there's no voice enrollment or audio being stored. There's also an upgraded custom Image Signal Processor that particularly helps video performance, especially in low-light conditions. The ISP performs more granular scene segmentation to understand regions and objects. That translates to the Pixel 10 recording 10-bit video by default at 1080p and 4K30. There's also motion deblur and improved Real Tone. On camera front, Tensor G5 powers Add Me, Auto Best Take, and 100x Pro Res Zoom. The latter is Google's largest Pixel Camera model at nearly 1 billion parameters. It's also the first diffusion model running in the app with the TPU helping play a big role, though the camera pipeline is using every part of the chip (CPU, DSP, ISP). For comparison, zoom on the Pixel 8 Pro's G3 was tens of thousands of parameters, while Pixel 9 Pro's's G4 was in the low millions. (Another example of G5's prowess is a 3-billion parameter multilingual model for real-time speech use cases.)

5 reasons why Google Tensor G5 is a game-changer for Pixel

Our vision for Pixel has always been to create a smartphone that's defined by smart, helpful features -- the kind that make your life easier. At the core of this is Tensor, Google's custom-designed mobile processor. When we introduced Tensor five years ago, it pushed the boundaries of what was possible on your Pixel phone through advanced machine learning. Now, in close partnership with Google Research teams, we're delivering the next chapter in AI innovations with the biggest upgrade to our custom silicon yet: Google Tensor G5. Here are five things to know about the new Tensor G5 chip: We've made enhancements across Tensor G5, including an up to 60% more powerful TPU and 34% faster on average CPU, making your Pixel more responsive for everyday use, like browsing the web, and the latest AI features. Tensor G5 is designed in the leading 3 nm process node from TSMC -- a manufacturing technology that allows us to pack more transistors into the chip so it's more powerful and efficient. New security hardware in Tensor G5 also protects Pixel even better throughout its lifecycle, from manufacturing to the moment it's in your hand. Thanks to our close partnership with Google DeepMind, Tensor G5 is the first chip to run the newest Gemini Nano model, unlocking many fully on-device generative AI experiences at launch (more on that below!). Not only does Tensor G5 run the highest-quality model, it does so faster and more efficiently than previous generations. In fact, with Tensor G5's higher performance and optimizations in the latest model, Gemini Nano runs 2.6x faster and 2x more efficiently for use cases like Pixel Screenshots and Recorder. Unique Pixel features like Magic Cue, Voice Translate, Call Notes with actions and Personal Journal bring you new kinds of proactive help all on-device thanks to Tensor G5. Tensor G5 with Gemini Nano also enhances features you already know and love, like Scam Detection and natural voice editing with Gboard. Pixel 10 Pro is the best smartphone camera for photography, supported by an all-new Image Signal Processor (ISP) in Tensor G5. Our ISP helps you capture the highest-quality videos on a Pixel every time, even in low lighting, with motion deblur and 10-bit video by default for 1080p and 4k30. Plus, it has improved Real Tone to seamlessly capture varying skin tones for everyone in the photo or video, so you can see each person in their most authentic light. In addition to image and video quality improvements, Tensor G5 powers a range of camera features including Add Me, Auto Best Take, and Pro Res Zoom, enabling the largest model ever to run in Pixel Camera to extend your zoom range to an amazing 100x with stunning detail. Pixel 10's Camera app is the first with C2PA Content Credentials built in. The full process takes place on-device on Tensor G5 and the Titan M2 security chip, where we create secure metadata within the image that documents its full journey. It achieved the highest security rating currently defined by C2PA. Even with all these new features packed in, every Pixel 10 phone has over 30 hours of battery life. That's enough power to cover you from the minute you wake up until long after you've gone to sleep.

Google confirms Pixel 10's new Tensor G5 chip is designed on the leading 3nm process from TSMC

TL;DR: Google's new Tensor G5 processor, built on TSMC's advanced 3nm process, powers the Pixel 10 series with up to 34% faster CPU and 60% improved TPU performance. Enhanced AI capabilities and a new Image Sensor Processor deliver superior responsiveness and high-quality video, marking Google's biggest chip upgrade yet. Google has some big upgrades going on under its custom Tensor G5 processor inside of its newly-announced Pixel 10 family of smartphones, including the new flagship Pixel 10 Pro Fold foldable smartphone. The company has posted a new blog explaining its new Google Tensor G5 processor being the "biggest upgrade yet" and delivering a major performance boost, with up to 34% more CPU performance. There's up to 60% more TPU performance. Mixed with the upgraded CPU performance, this makes the new Pixel smartphones more responsive with everyday use like web browsing, and the latest AI features (which are packed into Pixel 10). Google's new Tensor G5 processor is fabbed on TSMC's leading 3nm process node, allowing the company to cram in more transistors onto the chip, making it more powerful and efficient. There's an all-new Image Sensor Processor (ISP) inside of the new Tensor G5 chip, allowing higher-quality videos on a Pixel 10 smartphone, even in low lighting, with 10-bit video by default for 1080p, and 4K30. Google and TSMC have been working together on the new Tensor G5 chip on its 3nm process for the new Pixel 10 smartphones, with leaks of that from May 2024 in the link above. It was a big move for Google which had Samsung Electronics fab its Tensor G1, G2, G3, and recent G4 processors, leaving high-level executives at Samsung scratching their heads "intensively analyzing" how it lost Google to TSMC for Tensor chip production.

Google Has Announced Its First 3nm Chipset, The Tensor G5, Alongside The Pixel 10 Series; Company Claims A 34 Percent Average Performance Increase Over The Tensor G4, No RT Support & More

The Tensor G5 is not just Google's first SoC that has been mass produced on the 3nm manufacturing process, but is the first from the technology giant that is the product of TSMC's fabrication technology and not Samsung. Fortunately, the improved lithography is not the only change the company has introduced and thankfully, this chipset powers every Pixel 10 member. Here is everything you need to know about the latest silicon. The upgraded efficiency of the Tensor G5 means that the Pixel 10 lineup can enjoy a battery life increase of several hours, along with a significant boost in AI performance The CPU configuration of the Tensor G5 was not provided in detail, but Android Authority reports that during Google's briefing, the company mentioned that the new 3nm SoC features a '1 + 5 + 2' cluster, making it a significant change over last year's '1 + 3 + 4' specification. Unfortunately, we have yet to learn about the CPU clock speeds, but when the Pixel 10 Pro Fold was spotted in Geekbench 6, we reported that the fastest core operated at 3.78GHz, with the medium cores sporting a clock speed of 3.05GHz, and the slowest featuring speeds of 2.25GHz. Google claims that, on average, the Tensor G5 CPU is 34 percent faster than the Tensor G4, but failed to mention whether these improvements are in the single-core or multi-core department. Unfortunately, we do not expect Google's latest silicon to have the upper hand over the Snapdragon 8 Elite, Dimensity 9400+, or the A18 Pro in raw performance, and neither should you. As for the GPU, Google was somewhat mum on the details, but it did mention that the Tensor G5 was equipped with an updated graphics processor IP. Thankfully, previous reports stated that the CPU was paired with an Imagination Technologies DXT-48-1536 unit, though we still recommend readers to await confirmation. Unfortunately, Google has trimmed some features from the Tensor G5 GPU, as it lacks ray tracing support, giving the competition a stronger edge, at least in mobile gaming, which has a slow adoption rate. Of course, this unveiling would not be complete without Google's involvement of AI, with the company claiming that the chipset's TPU is up to 60 percent more powerful than the unit present in the Tensor G4. The more interesting addition of the Tensor G5 is that it features Google's 'Matryoshka' transformer model with two billion effective parameters. In short, the Gemini Nano model running on the chipset is 260 percent faster and twice as efficient as the Tensor G4 for tasks you will carry out on your Pixel 10 locally. Google also states that the token window has increased from 12,000 on the Tensor G4 to 32,000 on the Tensor G5, which is equivalent to 100 screenshots. Coming to the ISP, the improvements made help to substantially reduce blur from low-light videos, with 10-bit HDR added as the default for 1080p and 4K 30FPS video recording. As for the 8K local video recording, the Pixel 10 range has been omitted from this feature, but users can still take advantage of Google's cloud-based Video Boost. As for the 5G modem, the latest flagships still rely on the Exynos 5400 for wireless connectivity. If these upgrades appear attractive to you, you will be happy to learn that the Pixel 10 family is available to pre-order from Amazon, starting from $799, with the expected release date being August 28. Best of all, the online retailer is giving you a $100 gift card for the base model, with the freebie increasing to a $200 gift card for the Pixel 10 Pro or Pixel 10 Pro XL. Of course, we saved the best for last, and if you are willing to spend top dollar on the Pixel 10 Pro Fold, you will be rewarded with a mouth-watering $300 gift card, but you will have to fork over $1,799 for the flagship.

Google launches Tensor G5: Here's everything it does

Tensor G5 offers 34% faster CPU, 60% stronger TPU, enhanced privacy features Google has officially unveiled the Tensor G5 at the Made by Google 2025 event, positioning it as the most significant upgrade yet to its custom smartphone silicon. More than just a routine spec bump, the G5 is designed to be the cornerstone of an AI-first mobile experience, combining faster performance, smarter efficiency, and an expanded suite of on-device generative AI features. With the Pixel 10 lineup now built around it, Tensor G5 isn't just about raw power -- it's about redefining what a phone can do when artificial intelligence is embedded at its very core. Also read: Google Pixel 10 Pro vs Pixel 9 Pro: Price, camera, display, battery and more compared At the heart of the new Pixel series, Tensor G5 delivers performance leaps that make everyday tasks faster and more seamless. Manufactured using TSMC's advanced 3-nanometer process, the chip is more efficient and powerful than ever. Google claims the CPU is 34% faster on average compared to its predecessor, while the TPU, dedicated to AI processing, sees a remarkable 60% boost. For users, this translates into snappier app launches, smoother multitasking, and sustained performance without the heat issues that plagued earlier Tensor chips. Gaming and streaming also benefit from better thermal management, ensuring long sessions remain stable and responsive. Tensor G5 also marks a turning point for on-device generative AI. Co-developed with DeepMind, the chip is the first to fully support Gemini Nano, Google's lightweight AI model optimized for mobile. The numbers are striking: Gemini Nano runs 2.6 times faster and consumes half the energy compared to Tensor G4. Thanks to new techniques like the Matryoshka transformer architecture and Per-Layer Embedding, Tensor G5 can now scale AI performance dynamically, adjusting memory usage depending on the task. That allows for more ambitious processing, with the chip capable of handling up to 32,000 tokens of context, the equivalent of analyzing around one hundred screenshots or a month of emails in one go. The real benefit of this AI horsepower is how it translates into everyday Pixel features. Google has built over 20 new AI-driven tools into the Pixel 10 lineup, all powered by Tensor G5. From Magic Cue, which offers contextual prompts and assistance, to Voice Translate, which delivers live call translation in the speaker's own voice, the experience feels more proactive and natural. Other highlights include Call Notes with automatic action items, Personal Journal for smart reflections, Scam Detection during calls, and Gboard Smart Edit, which fine-tunes your writing with minimal effort. Crucially, all these experiences happen on-device, meaning they run instantly, without cloud latency, and with strong privacy guarantees, your personal data never has to leave the phone. Photography remains Pixel's crown jewel, and Tensor G5 pushes it even further. The new Image Signal Processor introduces 10-bit video capture by default at both 1080p and 4K30. Photography also sees improved motion deblur, smarter scene segmentation, and even more accurate Real Tone rendering, ensuring skin tones appear true-to-life. Also read: Google pokes fun at Apple throughout Pixel 10 launch On the Pixel 10 Pro models, Google debuts Pro Res Zoom, a feature powered by a diffusion model with nearly a billion parameters. This allows the device to achieve an extraordinary 100× hybrid zoom while maintaining clarity and detail. Combined with the chip's AI-driven scene analysis, Tensor G5 cements Pixel's reputation as the smartphone camera for enthusiasts and casual users alike. The camera also gains C2PA Content Credentials, embedding verifiable metadata to confirm when, where, and how an image was captured. Managed entirely by Tensor G5 in tandem with the Titan M2 security chip, this system helps ensure image authenticity in an age of AI-generated content. By shifting production to TSMC's 3-nm node, Google has delivered more performance per watt than any Tensor chip before it. This means longer battery life in everyday use, even when running intensive AI features. The design also enhances thermal stability, so the phone stays cooler under pressure. Security has been another area of focus. Tensor G5 is engineered with safeguards at every step, from manufacturing to everyday operation. Together with Titan M2, it ensures sensitive data remains protected while keeping AI processes private and local. While AI and CPU gains are clear, the GPU story is more complicated. Early reports suggest Tensor G5 uses a PowerVR-based GPU, which notably drops support for ray tracing. This could limit its appeal for gamers looking for cutting-edge graphics performance. Although casual gaming will still run smoothly, power users may find the Pixel 10 less competitive than rivals built on Qualcomm or Apple silicon when it comes to high-end gaming. With Tensor G5, Google has delivered far more than another generational upgrade. This is a chip built to showcase the company's vision of an AI-first smartphone, where features are proactive, personalized, and private. Whether it's translating conversations in real time, intelligently organizing your digital life, or capturing professional-grade photos, the Pixel 10 series powered by Tensor G5 redefines what users can expect from a modern phone. There are caveats, particularly for those who care most about gaming performance. But for the vast majority of users, and especially those invested in Google's ecosystem, Tensor G5 is the strongest case yet for sticking with Pixel.