Gemini hits 1B users, NVIDIA Nemotron 3.5 Lightning, and Hugging Face video models for daily use
TL;DR
Google's Gemini app crossed 1 billion monthly users while NVIDIA released efficient agent models and Hugging Face surfaced new image-to-video and medical tools that teams can test immediately.
What shipped
On 11 August 2026, releases from Google, NVIDIA, and Hugging Face showed AI shifting from chat interfaces to agent workflows and specialized media tools. Google reported the Gemini app reaching 1 billion monthly users. NVIDIA and open platforms added models and runtimes that let builders run longer agent tasks with lower cost.
Hugging Face trending
Hugging Face surfaced five trending models focused on video generation and medical reasoning. Lightricks and fal led with image-to-video and text-to-video options that run on the Hub. These releases give non-coders direct access to fine-tuning and inference without new infrastructure.
- •LTX-2.5 by Lightricks Lightricks released LTX-2.5, an image-to-video model now trending on Hugging Face. It supports fine-tuning and inference directly on the Hub for short video clips from still images.
- •MiniMax-H3-Realism-People-LoRA by fal fal published a text-to-video LoRA model trending on the Hub. Users can download it to generate realistic people footage from text prompts without extra setup.
- •fuse-1-Lite by Akahsizrr Akahsizrr added fuse-1-Lite, a text-generation model built with transformers and trending on Hugging Face. It offers quick fine-tuning for custom chat or summarization tasks.
- •Muse-Glimmer-30B-GGUF by unsloth unsloth released Muse-Glimmer-30B-GGUF, an image-text-to-text model now trending. The GGUF format lets developers run it locally for multimodal queries.
- •MedPixel medical model A new pixel-language model for medical reasoning and segmentation appeared on the Hub. It links clinical text with precise image regions for diagnostic workflows.
Vendor launches
Vercel, Google, and NVIDIA shipped updates that move AI from experiments into managed production. Google announced 1 billion Gemini users and a medical consultation prototype. NVIDIA added efficient agent models and power architecture changes for high-density racks.
- •Gemini app at 1 billion users Google reported the Gemini app reached 1 billion monthly users, the fastest growth in company history for any product.
- •AMIE clinical video system Google tested AMIE for real-time simulated medical video consultations that handle patient dialogue and visual assessment.
- •NVIDIA 800 VDC power architecture NVIDIA described an 800-volt DC power design that cuts losses when feeding dense GPU racks in AI factories.
- •Nemotron 3.5 Lightning and NeMo Switchyard NVIDIA released Nemotron 3.5 Lightning for long-running agent tasks and Switchyard for routing across models with lower latency.
- •NVIDIA local AI ecosystem push NVIDIA highlighted open models and tools that let developers run agents on laptops and workstations without cloud calls.
- •Google AI Professional Certificate expansion Google added vibe coding modules to its top Coursera certificate to teach non-coders how to direct AI agents.
- •Cybersecurity model shift note A Vercel post warned that open-weight models will soon match closed models in offensive security tasks, narrowing the current defender edge.
Replicate new models
ltx-2.5-fast on Replicate: Lightricks launched ltx-2.5-fast on Replicate for text-to-video and image-to-video with frame interpolation and synchronized audio up to 20 seconds at 1,080p.
Product Hunt picks
Two agent-focused tools appeared on Product Hunt. Spotify released an agentic development environment and Gitar introduced an AI reviewer that applies fixes automatically.
- •Xirp by Spotify Spotify launched Xirp, an agentic development environment that lets teams orchestrate code changes through guided agents.
- •Gitar AI code review Gitar released an AI reviewer that detects issues and submits fixes in the same pull request workflow.
Industry news
OpenAI and Anthropic advanced internal capabilities while Chrome strengthened account security. OpenAI enabled another large employee stock sale and added a Linux desktop app.
- •Anthropic math progress An unreleased Anthropic model made measurable headway on the long-standing Riemann hypothesis in mathematics.
- •ChatGPT desktop app for Linux OpenAI released a native ChatGPT desktop application for Linux users.
Other
Databricks, GitHub, NVIDIA, and Liquid AI released tools that support local agents and large codebases. Several models became available on OpenRouter for immediate testing.
- •Metals v2 from Databricks Databricks open-sourced Metals v2, a language server that handles multi-million-line Java and Scala codebases now written largely by agents.
- •GitHub agent role shift GitHub described how developers move from writing code to orchestrating agent delivery pipelines around it.
- •Nemotron 3.5 Lightning on OpenRouter NVIDIA's Nemotron 3.5 Lightning model is now routable on OpenRouter with 262K context at $0.10 per million input tokens.
- •Seed-2.0-Code on OpenRouter ByteDance added Seed-2.0-Code to OpenRouter with 262K context for coding tasks at $0.50 per million input tokens.
- •Daybreak Red and Blue on Bedrock OpenAI and AWS made Daybreak Red and Blue cyber-defense models available to eligible Amazon Bedrock customers.
- •LFM2.5-2.6B free on OpenRouter Liquid AI released LFM2.5-2.6B as a free 128K context reasoning model suited for RAG and agent workflows.
What this means for you
For Vibe Builders: You can now test LTX-2.5 and Nemotron 3.5 Lightning through Replicate and OpenRouter to generate video clips or run agent tasks without writing code. The Gemini app at 1 billion users and the expanded Google certificate show simple prompts already reach mainstream audiences. Start with one trending Hub model or the new OpenRouter endpoints this week to ship a short workflow before the next release cycle.
For Non-techies: The Gemini app now serves 1 billion people monthly and Chrome added stronger login protection, so daily tools are becoming more reliable for business tasks. NVIDIA and Lightricks models on Replicate let you create short videos from text or images without technical setup. Watch for the Linux ChatGPT app and managed user features on Vercel if you run teams that need simple account controls.
For Developers: Vercel added token observability and Enterprise Managed Users while NVIDIA released Nemotron 3.5 Lightning and an 800 VDC power design for dense racks. Databricks open-sourced Metals v2 for large agent-written codebases and OpenRouter exposed several new models with clear pricing. Benchmark the new agent runtimes against your current stack and test one OpenRouter endpoint before integrating into production pipelines.
What to watch next
Track whether OpenAI follows the Linux app with broader desktop features. Watch for production benchmarks on Nemotron 3.5 Lightning and any follow-up from Anthropic on the math result. Check Replicate and Hugging Face for new video or medical model updates in the next few days.
Harsh’s take
The day shows a clear split between consumer scale and production tooling. Google proved massive adoption is possible with a chat interface, yet most new releases target agents and specialized media that still require integration work. The defender advantage in security may shrink quickly as open models improve.
Vercel and NVIDIA focus on control and efficiency while smaller teams gain easier access through marketplaces and free tiers. The practical gap is no longer model quality but reliable routing and observability when agents run for hours.
Test one new endpoint from OpenRouter or Replicate against a current workflow this week and measure token cost plus failure rate before committing to any vendor stack.
by Harsh Desai
Sources
Hugging Face trending
- •LTX-2.5 by Lightricks trends on HuggingFace
- •MiniMax-H3-Realism-People-LoRA by fal trends on HuggingFace
- •fuse-1-Lite by Akahsizrr trends on HuggingFace
- •Muse-Glimmer-30B-GGUF by unsloth trends on HuggingFace
- •MedPixel: A Unified Pixel-Language Model for Medical Reasoning and Segmentation
Vendor launches
- •LaunchDarkly is now available on the Vercel Marketplace
- •Vercel Connect adds observability support
- •Everything hackable will get hacked
- •Enterprise Managed Users is now generally available
- •More than 1 billion people are using the Gemini app every month.
- •AMIE, our research medical AI system, demonstrates real-time clinical video consultation capabilities in a first-of-its-kind study.
- •Why Scaling AI Compute Performance Requires a New Power Architecture
- •NVIDIA Nemotron 3.5 Lightning and NeMo Switchyard Deliver Faster, Smarter, More Efficient Agentic AI
- •NVIDIA and Local AI Community Fuel Open Source Models and Intelligent Agents
- •Expanding the Google AI Professional Certificate with vibe coding
Replicate new models
Product Hunt picks
Industry news
- •Brad Lightcap, OpenAI’s longtime COO, is leaving to ‘start something new’
- •An unreleased Anthropic model made progress on one of math’s biggest unsolved problems
- •OpenAI lets employees cash out another $7 billion in stock
- •OpenAI launches ChatGPT desktop app for Linux
- •Accel closes oversubscribed $550M India fund within weeks, 19 months after its last
- •Chrome adopts what may be the best protection yet against account takeovers
Other
- •Open-sourcing Metals v2: Databricks’ Java and Scala language server for multi‑million line codebases
- •From coder to orchestrator: How agents shift the role of a developer
- •NVIDIA: Nemotron 3.5 Lightning added on OpenRouter
- •ByteDance Seed: Seed-2.0-Code added on OpenRouter
- •Accelerate cyber defense with OpenAI and AWS: Daybreak Red & Daybreak Blue now available to eligible customers on Amazon Bedrock
- •LiquidAI: LFM2.5-2.6B (free) now available on OpenRouter (128k context, $0.00/M in, $0.00/M out)
More AI news
- FeatureGeneral Catalyst leads $1.1B round for two-month-old River AI
General Catalyst led a $1.1 billion funding round for River AI. The startup was founded two months ago by xAI co-founder Igor Babuschkin.