Free, offline ChatGPT in your cellphone? Technically doable, mainly ineffective

[ad_1] <img class="e_Kg" decoding="async" loading="eager" sizes="(min-width: 64rem) 51.25rem, 80vw" title="gpt oss running on Android phone" srcset="https://www.androidauthority.com/wp-content/uploads/2025/08/gpt-oss-running-on-Android-phone.jpg 1920w, https://www.androidauthority.com/wp-content/uploads/2025/08/gpt-oss-running-on-Android-phone-64w-36h.jpg 64w, https://www.androidauthority.com/wp-content/uploads/2025/08/gpt-oss-running-on-Android-phone-1000w-563h.jpg 1000w, https://www.androidauthority.com/wp-content/uploads/2025/08/gpt-oss-running-on-Android-phone-1536w-864h.jpg 1536w, https://www.androidauthority.com/wp-content/uploads/2025/08/gpt-oss-running-on-Android-phone-675w-380h.jpg 675w, https://www.androidauthority.com/wp-content/uploads/2025/08/gpt-oss-running-on-Android-phone-300w-170h.jpg 300w, https://www.androidauthority.com/wp-content/uploads/2025/08/gpt-oss-running-on-Android-phone-1280w-720h.jpg 1280w, https://www.androidauthority.com/wp-content/uploads/2025/08/gpt-oss-running-on-Android-phone-840w-472h.jpg 840w" alt="gpt oss running on Android phone" src="https://www.androidauthority.com/wp-content/uploads/2025/08/gpt-oss-running-on-Android-phone.jpg"/> Robert Triggs / Android Authority

One other day, one other giant language mannequin, however information that OpenAI has launched its first open-weight fashions (gpt-oss) with Apache 2.0 licensing is a much bigger deal than most. Lastly, you'll be able to run a model of ChatGPT offline and totally free, giving builders and us informal AI lovers one other highly effective software to check out.

As typical, OpenAI makes some fairly massive claims about gpt-oss’s capabilities. The mannequin can apparently outperform o4-mini and scores fairly near its o3 mannequin — OpenAI’s cost-efficient and strongest reasoning fashions, respectively. Nonetheless, that gpt-oss mannequin is available in at a colossal 120 billion parameters, requiring some critical computing package to run. For you and me, although, there’s nonetheless a extremely performant 20 billion parameter mannequin accessible.

Are you able to now run ChatGPT offline and totally free? Properly, it relies upon.

In idea, the 20 billion parameter mannequin will run on a contemporary laptop computer or PC, offered you've got bountiful RAM and a strong CPU or GPU to crunch the numbers. Qualcomm even claims it’s enthusiastic about bringing gpt-oss to its compute platforms — assume PC fairly than cellular. Nonetheless, this does beg the query: Is it doable to now run ChatGPT fully offline and on-device, totally free, on a laptop computer and even your smartphone? Properly, it’s doable, however I wouldn’t advocate it.

What do you could run gpt-oss?

<img class="e_Kg" decoding="async" loading="lazy" sizes="(min-width: 64rem) 51.25rem, 80vw" title="NVIDIA GeForce RTX GPUs 3" srcset="https://www.androidauthority.com/wp-content/uploads/2023/01/Nvidia-GeForce-RTX-GPUs-3.jpg 1200w, https://www.androidauthority.com/wp-content/uploads/2023/01/Nvidia-GeForce-RTX-GPUs-3-675w-380h.jpg 675w, https://www.androidauthority.com/wp-content/uploads/2023/01/Nvidia-GeForce-RTX-GPUs-3-64w-36h.jpg 64w, https://www.androidauthority.com/wp-content/uploads/2023/01/Nvidia-GeForce-RTX-GPUs-3-1000w-563h.jpg 1000w, https://www.androidauthority.com/wp-content/uploads/2023/01/Nvidia-GeForce-RTX-GPUs-3-300w-170h.jpg 300w, https://www.androidauthority.com/wp-content/uploads/2023/01/Nvidia-GeForce-RTX-GPUs-3-840w-472h.jpg 840w" alt="NVIDIA GeForce RTX GPUs 3" src="https://www.androidauthority.com/wp-content/uploads/2023/01/Nvidia-GeForce-RTX-GPUs-3.jpg"/> Edgar Cervantes / Android Authority

Regardless of shrinking gpt-oss from 120 billion to twenty billion parameters for extra basic use, the official quantized mannequin nonetheless weighs in at a hefty 12.2GB. OpenAI specifies VRAM necessities of 16GB for the 20B mannequin and 80GB for the 120B mannequin. You want a machine able to holding your entire factor in reminiscence without delay to attain cheap efficiency, which places you firmly into NVIDIA RTX 4080 territory for ample devoted GPU reminiscence — hardly one thing all of us have entry to.

For PCs with a smaller GPU VRAM, you’ll need 16GB of system RAM in the event you can break up among the mannequin into GPU reminiscence, and ideally a GPU able to crunching FP4 precision knowledge. For every thing else, akin to typical laptops and smartphones, 16GB is absolutely reducing it fantastic as you want room for the OS and apps too. Primarily based on my expertise, 24GB RAM is required; my seventh Gen Floor Laptop computer, full with a Snapdragon X processor and 16GB RAM, labored at an admittedly fairly respectable 10 tokens per second, however barely held on even with each different software closed.

Regardless of it is smaller measurement, gpt-oss 20b nonetheless wants loads of RAM and a strong GPU to run easily.

After all, with 24 GB RAM being superb, the overwhelming majority of smartphones can't run it. Even AI leaders just like the Pixel 9 Professional XL and Galaxy S25 Extremely prime out at 16GB RAM, and never all of that’s accessible. Fortunately, my ROG Telephone 9 Professional has a colossal 24GB of RAM — sufficient to get me began.

The right way to run gpt-oss on a cellphone

<img class="e_Kg" decoding="async" loading="lazy" sizes="(min-width: 64rem) 51.25rem, 80vw" title="gpt oss prompt response" srcset="https://www.androidauthority.com/wp-content/uploads/2025/08/gpt-oss-prompt-response.jpg 1620w, https://www.androidauthority.com/wp-content/uploads/2025/08/gpt-oss-prompt-response-64w-43h.jpg 64w, https://www.androidauthority.com/wp-content/uploads/2025/08/gpt-oss-prompt-response-300w-200h.jpg 300w, https://www.androidauthority.com/wp-content/uploads/2025/08/gpt-oss-prompt-response-1000w-667h.jpg 1000w, https://www.androidauthority.com/wp-content/uploads/2025/08/gpt-oss-prompt-response-1296w-864h.jpg 1296w, https://www.androidauthority.com/wp-content/uploads/2025/08/gpt-oss-prompt-response-570w-380h.jpg 570w, https://www.androidauthority.com/wp-content/uploads/2025/08/gpt-oss-prompt-response-675w-450h.jpg 675w" alt="gpt oss prompt response" src="https://www.androidauthority.com/wp-content/uploads/2025/08/gpt-oss-prompt-response.jpg"/> Robert Triggs / Android Authority

For my first try and run gpt-oss on my Android smartphone, I turned to the rising choice of LLM apps that allow you to run offline fashions, together with PocketPal AI, LLaMA Chat, and LM Playground.

Nonetheless, these apps both didn’t have the mannequin accessible or couldn’t efficiently load the model downloaded manually, probably as a result of they’re based mostly on an older model of llama.cpp. As a substitute, I booted up a Debian partition on the ROG and put in Ollama to deal with loading and interacting with gpt-oss. If you wish to observe the steps, I did the identical with DeepSeek earlier within the 12 months. The downside is that efficiency isn’t fairly native, and there’s no {hardware} acceleration, that means you’re reliant on the cellphone’s CPU to do the heavy lifting.

So, how effectively does gpt-oss run on a top-tier Android smartphone? Barely is the beneficiant phrase I’d use. The ROG’s Snapdragon 8 Elite is likely to be highly effective, however it’s nowhere close to my laptop computer’s Snapdragon X, not to mention a devoted GPU for knowledge crunching.

gpt-oss can nearly run on a cellphone, however it's barely usable.

The token price (the speed at which textual content is generated on display screen) is barely satisfactory and positively slower than I can learn. I’d estimate it’s within the area of 2-3 tokens (a few phrase or so) per second. It’s not fully horrible for brief requests, however it’s agonising if you wish to do something extra complicated than say hiya. Sadly, the token price solely will get worse as the scale of your dialog will increase, finally taking a number of minutes to provide even a few paragraphs.

<img class="e_Kg" decoding="async" loading="lazy" sizes="(min-width: 64rem) 51.25rem, 80vw" title="High CPU use graph" srcset="https://www.androidauthority.com/wp-content/uploads/2025/08/High-CPU-use-graph.jpg 1920w, https://www.androidauthority.com/wp-content/uploads/2025/08/High-CPU-use-graph-64w-36h.jpg 64w, https://www.androidauthority.com/wp-content/uploads/2025/08/High-CPU-use-graph-1000w-563h.jpg 1000w, https://www.androidauthority.com/wp-content/uploads/2025/08/High-CPU-use-graph-1536w-864h.jpg 1536w, https://www.androidauthority.com/wp-content/uploads/2025/08/High-CPU-use-graph-675w-380h.jpg 675w, https://www.androidauthority.com/wp-content/uploads/2025/08/High-CPU-use-graph-300w-170h.jpg 300w, https://www.androidauthority.com/wp-content/uploads/2025/08/High-CPU-use-graph-1280w-720h.jpg 1280w, https://www.androidauthority.com/wp-content/uploads/2025/08/High-CPU-use-graph-840w-472h.jpg 840w" alt="High CPU use graph" src="https://www.androidauthority.com/wp-content/uploads/2025/08/High-CPU-use-graph.jpg"/> Robert Triggs / Android Authority

Clearly, cellular CPUs actually aren’t constructed for the sort of work, and positively not fashions approaching this measurement. The ROG is a nippy performer for my day by day workloads, however it was maxed out right here, inflicting seven of the eight CPU cores to run at 100% nearly continually, leading to a fairly uncomfortably scorching handset after only a few minutes of chat. Clock speeds shortly throttled, inflicting token speeds to fall additional. It’s not nice.

With the mannequin loaded, the cellphone’s 24GB was stretched as effectively, with the OS, background apps, and extra reminiscence required for the immediate and responses all vying for house. Once I wanted to flick out and in of apps, I might, however this introduced already sluggish token technology to a digital standstill.

One other spectacular mannequin, however not for telephones

<img class="e_Kg" decoding="async" loading="lazy" sizes="(min-width: 64rem) 51.25rem, 80vw" title="openai chatgpt o1 model logo header" srcset="https://www.androidauthority.com/wp-content/uploads/2024/09/openai-chatgpt-o1-model-logo-header-scaled.jpg 2560w, https://www.androidauthority.com/wp-content/uploads/2024/09/openai-chatgpt-o1-model-logo-header-1920w-1080h.jpg 1920w, https://www.androidauthority.com/wp-content/uploads/2024/09/openai-chatgpt-o1-model-logo-header-1536w-864h.jpg 1536w, https://www.androidauthority.com/wp-content/uploads/2024/09/openai-chatgpt-o1-model-logo-header-675w-379h.jpg 675w, https://www.androidauthority.com/wp-content/uploads/2024/09/openai-chatgpt-o1-model-logo-header-64w-36h.jpg 64w, https://www.androidauthority.com/wp-content/uploads/2024/09/openai-chatgpt-o1-model-logo-header-1000w-562h.jpg 1000w, https://www.androidauthority.com/wp-content/uploads/2024/09/openai-chatgpt-o1-model-logo-header-300w-170h.jpg 300w, https://www.androidauthority.com/wp-content/uploads/2024/09/openai-chatgpt-o1-model-logo-header-1280w-720h.jpg 1280w, https://www.androidauthority.com/wp-content/uploads/2024/09/openai-chatgpt-o1-model-logo-header-840w-472h.jpg 840w" alt="openai chatgpt o1 model logo header" src="https://www.androidauthority.com/wp-content/uploads/2024/09/openai-chatgpt-o1-model-logo-header-scaled.jpg"/> Calvin Wankhede / Android Authority

Operating gpt-oss in your smartphone is just about out of the query, even in case you have an enormous pool of RAM to load it up. Exterior fashions aimed primarily on the developer neighborhood don’t help cellular NPUs and GPUs. The one method round that impediment is for builders to leverage proprietary SDKs like Qualcomm’s AI SDK or Apple’s Core ML, which received’t occur for this form of use case.

Nonetheless, I used to be decided not to surrender and tried gpt-oss on my growing older PC, geared up with a GTX1070 and 24GB RAM. The outcomes had been undoubtedly higher, at round 4 to 5 tokens per second, however nonetheless slower than my Snapdragon X laptop computer working simply on the CPU — yikes.

In each instances, the 20b parameter model of gpt-oss actually appears spectacular (after ready some time), due to its configurable chain of reasoning that lets the mannequin “assume” for longer to assist remedy extra complicated issues. In comparison with free choices like Google’s Gemini 2.5 Flash, gpt-oss is the extra succesful drawback solver due to its use of chain-of-thought, very like DeepSeek R1, which is all of the extra spectacular given it’s free. Nonetheless, it’s nonetheless not as highly effective because the mightier and costlier cloud-based fashions — and positively doesn’t run wherever close to as quick on any shopper devices I personal.

Nonetheless, superior reasoning within the palm of your hand, with out the fee, safety considerations, or community compromises of right this moment’s subscription fashions, is the AI future I believe laptops and smartphones ought to actually intention for. There’s clearly an extended technique to go, particularly on the subject of mainstream {hardware} acceleration, however as fashions grow to be each smarter and smaller, that future feels more and more tangible.

A couple of of my flagship smartphones have confirmed moderately adept at working smaller 8 billion parameter fashions like Qwen 2.5 and Llama 3, with surprisingly fast and highly effective outcomes. If we ever see a equally speedy model of gpt-oss, I’d be far more excited.