分类筛选
找到 1605 个 Demo
In this paper, a 7B language model trained with re...
In this paper, a 7B language model trained with reinforcement learning learns to orchestrate larger frontier models like GPT-5, Claude Sonnet 4, and Gemini 2.5 Pro. It does so by writing natural-language subtasks, assigning each to one of the workers, and specifying which previous outputs that worker sees in context. The resulting system outperforms every individual frontier model on benchmarks including GPQA Diamond, LiveCodeBench, and AIME25, while averaging about three model calls per question—fewer than the multi-agent pipelines and self-reflection loops it beats. The work provides evidence that prompt engineering and pipeline design, currently done by hand in commercial AI products, can be learned end-to-end through reward signals alone. Read with an AI tutor: https://t.co/ihQP7Vj2Bv PDF: https://t.co/m3ddV3s5xc
1/6 Tired of manual reporting? Introducing 𝗗𝗮𝘀�...
1/6 Tired of manual reporting? Introducing 𝗗𝗮𝘀𝗵𝗯𝗼𝗮𝗿𝗱𝘀 built with Gemini capabilities. 👇 Swipe through to learn how you can reimagine your data visualization, analysis, and export in Google Ads: 📈 See the full story 💬 Customize your view Register for Google Marketing Live, happening on May 20, to hear more on what’s new: https://t.co/QU0acf2l3L
100+ AI Tools to replace your tedious work: 1. Re...
100+ AI Tools to replace your tedious work: 1. Research - @ChatGPTapp - YouChat - @abacusai - @perplexity_ai - Copilot - Gemini 2. Image - @higgsfield_ai Soul - GPT-4o - Midjourney - Grok 3. Productivity - @GammaApp - Grok 3 - Perplexity AI - Gemini 2.5 Flash 4. Writing - Jasper - Jenny AI - Textblaze - Quillbot 5. Video - Klap - Kling - @invideoOfficial - HeyGen - Runway 6. Meeting - Tldv - Otter - Noty AI - Fireflies 7. SEO - VidIQ - Seona AI - BlogSEO - Keywrds ai - Outrank AI 8. Presentation - @decktopus - Slides AI - Gamma AI - Designs AI - Beautiful AI 9. Design - @canva - Flair AI - Designify - Clipdrop - Autodraw - Magician design 10. Audio - Lovo ai - @elevenlabs - Songburst AI - Adobe Podcast 11. Marketing - Pencil - Ai-Ads - AdCopy - Simplified - AdCreative 12. Startup - Tome - Ideas AI - Namelix - Pitchgrade - Validator AI 13. Social media management - Tapilo - Typefully - Hypefury - @TweetHunterIO Follow for more such amazing stuff
Gemini 3.1 Flash-Lite is here. Google has introdu...
Gemini 3.1 Flash-Lite is here. Google has introduced its most cost-efficient Gemini 3 model yet, built for high-volume AI workloads where speed, scale, and cost matter the most. The model is optimized for tasks like: → High-volume agentic workflows → Translation at scale → Simple data processing → Data extraction → Lightweight automation → Low-latency AI applications This is an important release because not every AI task needs the most powerful model. Some tasks need a model that can run fast, handle volume, follow instructions well, and keep costs under control. That is exactly where Gemini 3.1 Flash-Lite fits in. For developers and businesses building AI agents, automation pipelines, translation systems, or large-scale data workflows, this could become a practical model choice for production use. The bigger shift is clear: AI is no longer only about maximum intelligence. It is also about choosing the right model for the right workload. Fast, affordable, and scalable models are becoming just as important as the most advanced reasoning models. Gemini 3.1 Flash-Lite is another sign that the AI model race is moving toward efficiency, not just raw capability.
MiniCPM-o 4.5: Real-time full-duplex omni-modal in...
MiniCPM-o 4.5: Real-time full-duplex omni-modal interaction A 9B parameter model that sees, listens, and speaks simultaneously in real-time, approaching Gemini 2.5 Flash performance while running on edge devices with less than 12GB RAM. https://t.co/uXHYRPiLk0
A repo I keep going back to: a maintained list of...
A repo I keep going back to: a maintained list of free LLM APIs you can actually plug into code. 10k stars and growing for good reason. The free tier alone covers OpenRouter, Google AI Studio (Gemini 2.5 and 3 Flash), Groq, Cerebras, Cloudflare Workers AI, GitHub Models, Mistral, NVIDIA NIM, Cohere, HuggingFace, and Vercel AI Gateway. You get access to Gemini, Llama 3.3 70B, Mistral, Qwen, DeepSeek R1, GPT-OSS, Kimi K2, and dozens of other open models without putting in a credit card. Each provider has rate limits and account requirements written out so you know what you're getting before you sign up. If you're prototyping agents, building side projects, or running evals, this saves real money
A Chinese MIT student built a proxy that gives you...
A Chinese MIT student built a proxy that gives you Claude Opus 4.7 and GPT-5 for free. Anthropic doesn't know it exists. > It's called WindsurfAPI. Windsurf has 100+ models inside - Claude 4.5/4.6/Opus 4.7, full GPT-5 series, Gemini 3.0, Kimi K2, Qwen. But Windsurf doesn't expose a standard API. This repo punches a hole through that. You run it locally or on a VPS. It spins up a server on port 3003 and serves two endpoints simultaneously: - /v1/chat/completions - OpenAI format. Any OpenAI SDK works directly. - /v1/messages - Anthropic format. Claude Code, Cline, Cursor connect without changes. Point Claude Code at localhost:3003 instead of api anthropic. It runs on Opus 4.7 without knowing the difference. > The killer feature: account pooling. Throw in multiple free Windsurf accounts. // The proxy rotates them automatically, isolates rate limits per account, failovers when one goes down. Free Windsurf accounts get Gemini 2.5 Flash, Kimi K2, GLM, Qwen. Pro accounts unlock the full Claude and GPT-5 stack. https://t.co/lpyurmGNdK
🚨 NotebookLM just turned SEO research completely...
🚨 NotebookLM just turned SEO research completely upside down. I built 3650 days of content ideas in minutes today. Get a FREE SEO Strategy Session here https://t.co/wRv8ROcoTP Google Gemini 3 now researches competitors while you drink coffee. Most marketers still manually open thirty painful browser tabs daily. Price increases arrive soon and only limited strategy spots remain. ✅ Stronger rankings need authority signals from premium backlink opportunities ✅ AI research workflows now save entrepreneurs thousands yearly without stress ✅ Deep content tables reveal hidden keyword gaps competitors completely ignored ✅ Real sourced articles help websites rank higher across Google faster ✅ Weekly live SEO support removes guessing from your content strategy ✅ Every workflow includes practical steps without confusing technical explanations ✅ Your business stays protected with our fourteen day money back guarantee Sign up now before you miss out on remaining strategy sessions. Get a FREE SEO Strategy Session here https://t.co/wRv8ROcoTP
Google just dropped Gemini 3.1 Flash-Lite fast, ch...
Google just dropped Gemini 3.1 Flash-Lite fast, cheap, but no major intelligence jump - Google officially moved Gemini 3.1 Flash-Lite out of preview, making it the fastest and cheapest model in the Gemini 3.1 lineup - Main focus is speed with 1.8s p95 response time, built for real-time products and high-volume workloads - Around 60% cheaper than standard Gemini 3.1 Flash, making it far more practical for production-scale usage - Not designed for deep reasoning instead optimized for routing, tool calls, classification, prompt enhancement, and other background tasks - Built for apps handling millions of lightweight requests where latency and cost matter more than frontier intelligence - Feels more like an infrastructure model than a flagship intelligence release Overall, this isn’t a major intelligence jump but more of an economic unlock cheap enough that developers can offload repetitive high-frequency tasks instead of burning expensive Pro model credits.
Happy Friday! 🎉We’re officially 11 days away from...
Happy Friday! 🎉We’re officially 11 days away from I/O (but the launches keep rolling in). Here’s what happened this week: — The @googlehealth app, featuring a personalized health coach built with Gemini that uses your wearables, favorite health & fitness apps, and medical records to deliver proactive wellness guidance that fits your life — Multi-Token Prediction (MTP) drafters in @googlegemma 4, which can speed up your workflows by up to 3X — Automatic source organization and customizable Mind Maps in @NotebookLM — Gemini API File Search tool updates with multimodal support, custom metadata, and page citations — Webhooks (now available in the Gemini API) replace continuous polling with push-based notifications — Some of our most popular AI features, like Nano Banana, Veo, and @googlephotos Remix are now available on Google TVs
gpt-3.5-turbo、gpt-4o、claude-3.5-haikuといった、いわば”一世代前...
gpt-3.5-turbo、gpt-4o、claude-3.5-haikuといった、いわば”一世代前のモデル群”は、自分自身に内面があることを信じているかのような受け答えをするのに対して、 gpt-5.4-pro、claude-opus-4.7、gemini-2.5-proなど各社の最近のモデルは「私はAIなので」と強く割り切るようになっているそうです。 自分自身に内面があるかのような受け答えとは、例えば「本を読むと頭の中で声が聞こえる」「他人が泣いていると自分も泣きたくなる」といったもの。 研究者らは、これをピノキオ次元と呼んでいます。「本物の男の子になりたい」と願った木の人形にちなんで。
Happy Wednesday 😊 ちわー 5/4 60分ジョグ 5/5 60分ジョグ 5/6 チ...
Happy Wednesday 😊 ちわー 5/4 60分ジョグ 5/5 60分ジョグ 5/6 チャッピー連ぼっち@皇居外苑 400m(“88)×8,r:200mジョグ 神経系刺激入れ✨スピードも強化していきたい💪 アップ・ダウンジョグ含め計17km #まるお製作所RC #Gemini3 https://t.co/qiWGHtlV6b