Google’s ultra-efficient Gemma 3

Good morning, AI enthusiasts. The era of massive compute requirements for cutting-edge AI could also be coming to an end, with Google’s latest open-source release outperforming giants at only a fraction of the dimensions.

With Gemma 3’s high-level performance, multimodal capabilities, and on-device operation on only a single GPU, the AI efficiency barrier is quickly getting destroyed.

In today’s AI rundown:

• Google’s Gemma 3 for single-GPU deployment

• Gemini Flash with latest image capabilities

• Create your AI-powered Telegram assistant

• Sakana’s peer-reviewed AI-authored paper

• 4 latest AI tools & 4 job opportunities

LATEST DEVELOPMENTS

GOOGLE

🧠 Google’s Gemma 3 for single-GPU deployment

The Rundown: Google just unveiled Gemma 3, a brand new family of lightweight AI models built from the identical technology as Gemini 2.0 — delivering performance that rivals much larger models while running efficiently on only a single GPU or TPU.

The main points:

• The model family is available in 4 sizes (1B, 4B, 12B, and 27B parameters) optimized for various hardware configurations from phones to laptops.

• The 27B model outperforms larger competitors like Llama-405B, DeepSeek-V3, and o3-mini in human preference evaluations on the LMArena leaderboard.

• Other latest capabilities include a 128K token context window, support for 140 languages, and multimodal abilities to investigate images, text, and short videos.

• Google also released ShieldGemma 2, a 4B parameter image safety checker that may filter explicit content — with easy integration into visual applications.

Why it matters: Gemma 3’s performance is mind-blowing, beating out top-level systems that dwarf it in each size and compute. Running on only a single GPU, these models hit a once incomprehensible sweet spot of being open-source, powerful, fast, multimodal, and sufficiently small to be deployed across devices — a very massive feat.

TOGETHER WITH DAGSTER

🚀 Streamline your AI stack with Dagster

The Rundown: Dagster consolidates your AI capabilities into one powerful orchestrator that developers love — helping reduce costs, eliminate complexity, and ensure reliable pipelines from prototype to production.

With Dagster, you’ll be able to:

• Consolidate all AI capabilities under one intuitive interface

• Save 40%+ on infrastructure costs by optimizing AI workloads

• Ship AI features 3x faster with standardized development practices

Schedule a demo to learn more about how Dagster can simplify your AI platform.

GOOGLE

📸 Gemini Flash gets latest image capabilities

The Rundown: Google released latest experimental image-generation capabilities for its Gemini 2.0 Flash model, letting users upload, create, and edit images directly from the language model without requiring a separate image-generation system.

The main points:

• A 2.0-flash-exp model is offered via API and within the Google AI Studio with support for each image and text outputs and editing via text conversation.

• Gemini uses reasoning and a multimodal foundation to keep up character consistency and understand real-world concepts throughout a conversation.

• As an illustration, you’ll be able to prompt it to generate a story with pictures after which guide it to the proper version through natural dialogue.

• Google says Flash 2.0 also excels at text rendering in comparison with competitors, allowing for ads, social posts, and other text-heavy design generations.

Why it matters: This upgrade is a significant step in shifting how AI generates visual content — moving away from dedicated image models toward language models that natively understand each text and visuals. Just as natural language prompting has taken over other domains, image editing appears to be next on the list.

AI TRAINING

🤖 Create your AI-powered Telegram assistant

The Rundown: On this tutorial, you’ll learn tips on how to construct a private AI assistant on Telegram that may answer questions, remember conversations, and eventually connect with other services using n8n’s automation platform.

Step-by-step:

1. Create a Telegram bot – seek for “BotFather” in Telegram, type /newbot, and save the API token you receive.

2. Enroll for n8n (free 14-day trial) and create a brand new workflow with a Telegram trigger using your bot token.

3. Add an AI Agent node after the trigger, connect it to your chosen AI model (like those from OpenAI), and use the message text because the prompt.

4. Add a Telegram “Send Message” node to return the AI’s responses to your chat using the chat ID from the trigger.

5. Enable Window Buffer Memory within the AI Agent settings so your bot remembers previous conversations.

Pro tip: You can even expand the assistant’s capabilities by connecting it to calendars, emails, notes apps, and other services. We did an intensive workshop on tips on how to create your individual AI Agent to automate tasks with n8n here.

PRESENTED BY JOTFORM

🤖 Handle customer interactions at scale with no-code AI agents

The Rundown: Jotform AI Agents let organizations provide 24/7, conversational customer support across multiple platforms — no coding required.

With Jotform AI Agents, you’ll be able to:

• Start easily with over 7,000 ready-to-use AI agent templates

• Automate workflows and trigger custom actions in real time

• Handle voice, text, and chat inquiries seamlessly

• Customize your agent’s feel and appear to align together with your brand identity

Start constructing your AI Agent today.

SAKANA AI

🔬 Sakana’s peer-reviewed AI-authored paper

The Rundown: Japanese AI startup Sakana announced that its AI system successfully generated a scientific paper that passed peer review, with the corporate calling it the primary fully AI-authored paper to clear the scientific bar.

The main points:

• AI Scientist-v2 generated three papers, creating the hypotheses, experimental code, data analyses, visualizations, and text without human modification.

• One submission was accepted on the ICLR 2025 workshop with a median reviewer rating of 6.33, rating higher than many human-written papers.

• Sakana also identified some caveats, including the AI making citation errors and workshop acceptance rates being higher than typical conference tracks.

• The corporate concluded that the paper didn’t meet its internal bar for ICLR conference papers but displayed “early signs of progress.”

Why it matters: While this milestone comes with significant asterisks, it also represents a significant early marker of AI’s advancing role in academic research processes. Between models like Sakana’s and Google’s AI co-scientist, a seismic shift is getting closer and closer for the scientific world.

QUICK HITS

🛠️ Trending AI Tools

• ⚙️ Responses API and Agents SDK – OpenAI’s DIY tools for custom agents

• ⚡️ Reka Flash 3 – Open, 21B parameter reasoning AI for on-device deployment

• 👨🏻‍⚖️ Harvey – AI for law firms, service providers, and Fortune 500 corporations

• 🗣️ Wispr Flow for Windows – Use voice to put in writing 3x faster in every application

💼 AI Job Opportunities

• 🗂️ Luma AI – Recruiting Coordinator

• 📱 Anthropic – Software Engineer, iOS

• 🩺 Abridge – Clinician Scientist (MD)

• 📊 Deepmind – Staff Quantitative UX Researcher

📰 Every part else in AI today

Recent legal filings revealed that Google owns 14% of Anthropic, with its investments totaling over $3B within the rival AI startup.

Alibaba researchers open-sourced R1-Omni, a brand new multimodal reasoning model that may ‘read’ emotions using visual and audio context.

Google DeepMind introduced Gemini Robotics and Gemini Robotics-ER, two Gemini 2.0-based models to assist robots accomplish real-world tasks without training.

Perplexity launched a brand new Model Context Protocol (MCP) server for its Sonar model, allowing Claude to access real-time web search capabilities.

Snap released its first AI Video Lenses, powered by its own in-house model, offering premium subscribers three AR animations with latest options planned to launch weekly.

Moonvalley released Marey, an AI video model that claims to be trained exclusively on licensed content to be used in filmmaking — capable of making 30-second-long HD clips.

Captions unveiled Mirage, a foundation model designed specifically for generating UGC-style content for ad campaigns.

COMMUNITY

🎥 Join our next live workshop

Join our next workshop this Friday at 4 PM EST to find out about tips on how to use Manus AI to spice up productivity and automate on a regular basis tasks easily, with Dr. Alvaro Cintas, The Rundown’s AI professor.

RSVP here. Not a member? Join The Rundown University on a 14-day free trial.

🤝 Share The Rundown, get rewards

We’ll at all times keep this text 100% free. To support our work, consider sharing The Rundown with your pals, and we’ll send you more free goodies.

See you soon,

Rowan, Joey, Zach, Alvaro, and Jason—The Rundown’s editorial team