on-device AI

12 stories tagged on-device AI.

MiniCPM5-2B: A 2.5B Model With 131K Context for On-Device AI

OpenBMB's MiniCPM5-2B averages 53.9 across 34 benchmarks and targets on-device use. What the model-card numbers show, and what they leave untested.

Same Model, 30% to 95%: Why the Harness Beats the Weights

Yuki Okonkwo2 days ago

AI. Yuki Okonkwo2 days ago

Same Model, 30% to 95%: Why the Harness Beats the Weights

YC Paper Club argues the scaffolding around an LLM, not the weights, drives results. Prime Agent, OpenJarvis, and QM show how far the wrapper can go.

Superwhisper's S1-mini Cleans Up ASR Transcripts On-Device

Yuki Okonkwo3 weeks ago

AI. Yuki Okonkwo3 weeks ago

Superwhisper's S1-mini Cleans Up ASR Transcripts On-Device

Superwhisper's S1-mini is a 462 MB open-weights model that strips fillers and fixes self-corrections in speech-to-text output—entirely on your device.

A man in a pink shirt stands beside glowing circuit board graphics with red and green neon lines, alongside text about…

Apple's Neural Engine: Specialized Chips vs. Data Centers

Yuki Okonkwo4 weeks ago

AI. Yuki Okonkwo4 weeks ago

Apple's Neural Engine: Specialized Chips vs. Data Centers

Apple's Neural Engine isn't an AI brain—it's a multiplication machine. Here's why that distinction matters for how businesses think about AI compute costs.

Gemini Nano Gets Faster on Pixel Without Retraining

Marcus Chen-Ramirez2 months ago

AI. Marcus Chen-Ramirez2 months ago

Gemini Nano Gets Faster on Pixel Without Retraining

Google's frozen Multi-Token Prediction retrofits speed gains onto existing Gemini Nano models—no retraining needed. Here's what that means for on-device AI.

Person wearing glasses against Earth backdrop with AI model comparison chart showing Qwen and Llama parameters, AI Engineer…

When Small AI Models Beat Frontier Ones on Your Tasks

Dev Kapoor2 months ago

AI. Dev Kapoor2 months ago

When Small AI Models Beat Frontier Ones on Your Tasks

RL Nabors walks through a real eval framework for replacing frontier model calls with local SLMs—and the results are more nuanced than the pitch suggests.

Man with glasses presenting AI research papers about text-guided image editing and Google DeepMind technology against dark…

Text Diffusion AI: Speed, Privacy, and Ambient Risk

Rachel "Rach" Kovacs3 months ago

AI. Rachel "Rach" Kovacs3 months ago

Text Diffusion AI: Speed, Privacy, and Ambient Risk

Google DeepMind's text diffusion model generates AI responses differently—and faster. Here's what that architectural shift means for privacy and everyday users.

Woman with brown hair in front of AI architecture diagrams showing attention mechanisms and MoE layers, with AI Engineer…

Google's Gemma 4 Makes Powerful AI Run on Your Phone

Yuki Okonkwo5 months ago

AI. Yuki Okonkwo5 months ago

Google's Gemma 4 Makes Powerful AI Run on Your Phone

Gemma 4 brings multimodal AI models to phones and laptops with clever architecture tricks that make 5B parameters perform like much larger models.

Phone displaying Google AI Edge Gallery interface with Gemma 4 options, alongside text promoting local AI model running…

Google Just Made Running LLMs on Your Phone Actually Simple

Dev Kapoor5 months ago

AI. Dev Kapoor5 months ago

Google Just Made Running LLMs on Your Phone Actually Simple

Google's AI Edge Gallery lets anyone run large language models locally on their phone—no developer account, no cloud, no data sharing. Here's what that means.

Bold red and blue text announcing Google's 100x release with "Gemma 4" highlighted in a white rounded box below

Google's Gemma 4: Running Frontier AI on Your Phone

Mike Sullivan5 months ago

AI. Mike Sullivan5 months ago

Google's Gemma 4: Running Frontier AI on Your Phone

Google's Gemma 4 brings frontier-level AI to consumer devices. Free, open-source, and offline-capable—but does it deliver on the promise?

Quinn 3.5 Runs AI Models On Your Phone Without Internet

Rachel "Rach" Kovacs6 months ago

AI. Rachel "Rach" Kovacs6 months ago

Quinn 3.5 Runs AI Models On Your Phone Without Internet

The Qwen 3.5 AI model runs entirely on your iPhone with zero internet connection. We tested how well local AI works when privacy actually matters.

Blue gradient background with "THIS IS INSANE!" headline and Google AI Edge branding showing four key features: on-device,…

Google's AI Edge: Revolution or Just Hype?

Marcus Chen-Ramirez8 months ago

AI. Marcus Chen-Ramirez8 months ago

Google's AI Edge: Revolution or Just Hype?

Google AI Edge lets AI models run on phones sans cloud, sparking debates on privacy and performance.