Peter Gostev’s Post

The most impressive achievement of DeepSeek R1, with the added benefit of being open-source, is that it eliminates the need for an average user to spend $20 a month and get plus plans of OpenAI, Claude, or Gemini. It does everything that the paid plans can do and more. DeepThink R1 is that good. And the way it thinks and talks us through the thinking process is even better than the final results. #DeepSeekAI #OpenSourceLLM #AIAccessibility

Google Gemini Gems.... No, I'm not giving you an instruction on something to search for, I'm talking about Google Gemini's latest feature update called Gems. They are a way to save your prompts, which initially sounds similar to OpenAI's GPTs, Microsoft Copilots, Anthropic Projects, etc. I was beyond excited, as we at HQ view this as the single most important thing to unlock self-service GenAI at your company. This came out about 2-3 weeks ago and I thought I has missed something big. Until I started to do more research. Gemini Gems are not able to be shared with coworkers and you cannot save knowledge files (e.g. docs, xls, pdfs, etc). Wait what?!? Yes you heard me right. Gems are purely just a way to save a simple prompt (and nothing else). Maybe this is why I didn't hear a lot of news about it when it first hit the market. What do you think? Are google gems worth it? Or are you sticking with your current GenAI prompting platform? Lemme know! [This post was Human Generated, Human Approved] #HQforAI #GPTs #OpenAI #GoogleGemini

Gemma 2 released! Google just released the next iteration of its open LLM! Gemma 2 comes in two sizes, 9B & 27B, trained on 13T tokens. Gemma 2 27B approaches Meta Llama 3 70B performance! First Chatbot Arena evals place Gemma2 27B around Anthropic Claude 3 Sonnet, Llama 3 70B, and OpenAI GPT-4. 🤯 What's new with Gemma 2: 🧮 9B & 27B Instruction and base version with 8192 context window 🔠 Trained on 13T tokens (27B) and 8T tokens (9B) 🆕 Sliding window attention, logit soft-capping and Grouped-Query Attention (GQA) 🥇 9B scores 71.3 MMLU; 52.8 AGIEval; 40.2 HumanEval 🏆 27B scores 75.2 MMLU; 55.1 AGIEval; 51.8 HumanEval ✅ Commercial use allowed 🧬 Used SFT, Distillation, RLHF & Model Merging. 🧠 Trained on Google TPUv5e 🤗 Available on Hugging Face 🔜 1-click deployment to Google Cloud from Hugging Face

Big News from OpenAI! OpenAI just launched the Realtime API. This means developers can now build super-fast speech-to-speech features in their apps. Think of creating apps where you can talk to an AI in real-time or have voice commands without delays. This makes using voice-based tech even smoother and faster. ______________ Follow me Muhammad Hassaan for more!

The biggest advantage of LLMs with Vision capabilities is that you no longer need to spend as much time on screen as you do now. You can start writing your thoughts, prompts and drawing your wireframes on a piece of paper and magically see them turn into something digital. Imagine using a Rocketbook and integrate it with a Multi-Modal LLM like OpenAI. #LLM #OpenAI

🚀 Call 200+ LLMs with OpenAI SDK!🚀 One of the features Keywords AI (YC W24) users most love is our compatibility with OpenAI SDK. ✅ Call 200+ LLMs using the same format ✅ Support extra reliability settings ✅ Integrate within minutes

I'm excited to share that we just released Google Gemini Flash 2.0 with thinking capabilities on AI Drive! 🚀 I compare it side-by-side with OpenAI's O1, and the results are fascinating: 🔷 Speed & Reasoning: Flash starts incredibly fast with real-time streaming You can actually see the model's thinking process as it works O1 provides complete responses in one batch 🔷 Key Differences: Flash 2.0 provides detailed reasoning tokens More cost-effective solution At first glance, both models deliver high-quality analysis with different approaches, but Flash tends to provide longer, more detailed responses 🚨 Sign up for a free trial account now at https://lnkd.in/eTfB_2Gi to try it out and let me know in the comments which one you prefer! #ArtificialIntelligence #ProductivityTools #AIInnovation #DocumentManagement #GoogleGemini #OpenAI

Market share of OpenAI is going down! Interesting trend of market share of OpenAI (2023 vs 2024). Their share has come down from 50% to 34% while Anthropic has gained market share from 12% to 24% and Google has jumped from 7% to 12%.

** Weather App with Gemini GPT AI Gemini GPT is a multimodality artificial intelligence assistant that saves you time in changing and optimizing code and content. All of that support by #flutterui material widgets. #flutterMobileDev  #GeminiGPT #ArtificialIntelligence #Dart3 #GenerativeAI #Google #Authentication #Firebase #FirebaseAuth #Login #Register #MVVM #Apis #VisualCrossing #Hours #Daily #GetX #MobileAppDeveloper

There was a time when I underestimated Google's forays into AI, given their earlier inconsistencies, but now I must admit my oversight. The recent capabilities of Gemini have proven to be nothing short of extraordinary, quickly becoming my favorite 💖! Following the triumph of Gemini 1.5 Pro in topping the lmsys rankings 🏆, the Gemini 1.5 Flash continues to impress with several notable enhancements: 1. Significant Price Reductions 💰: Starting on August 12th, the operational costs for Gemini 1.5 Flash will see a dramatic decrease. Input costs are reduced by 78% (now $0.075 per million tokens), and output costs are down by 71% (now $0.3 per million tokens). This pricing advantage applies to prompts with a maximum of 128K tokens and extends to higher token tiers and caching services. 2. Multilingual Support 🌍: Now extends to over 100 languages. 3. Seamless Integration 🔄: Workspace users can access Google AI Studio directly. 4. Open Fine-Tuning 🛠️: All developers can now utilize the fine-tuning functionalities. 5. PDF Analysis 📄: Enhanced capabilities for understanding text and imagery within PDFs. 6. Upgraded AI Studio 💻: A refined user experience with significantly faster loading times. If OpenAI doesn’t accelerate its pace, it risks being overtaken quickly! 🚀 #AI #OpenAI #Gemini #GeminiFlash