Reka

Meet Reka Edge – Our next-generation vision language model for physical AI. Uses 3x fewer input tokens and achieves 65% faster throughput compared to leading 8B models. Image understanding, video analysis, object detection, and tool use. Built for Action. Fast enough for production, deployable anywhere. Read more: reka.ai/news/reka-edge…

I put together a short tutorial that explains step-by-step how to get the brand new Reka Edge to run locally. You are only a few steps away from asking an #AI what's in any image or video 👉 link.reka.ai/reka-edge-dev 👀 #visionai #edgeai @RekaAILabs @huggingface

AI is moving from our screens into our physical reality. 🙅 But in a busy warehouse or security system, a two-second delay is a lifetime. Imagine having robot stop to think while doing a task 😅 🚀 At Reka, we believe intelligence should be fast, lean, and deployable anywhere, and we've spent the last few months solving the "Physical AI" bottleneck We’ve just launched Reka Edge. It’s not just faster; it’s built for the Physical World. Hear from Jacky Mok @aspheris, our Head of Applied AI, on how we're rethinking visual reasoning

Impressive 7B multimodal vision language model 🔥 Available on @huggingface 👉huggingface.co/RekaAI/reka-ed…

98ms time to first token (faster than human visual reaction time), built for agentic workflows. 65% faster throughput compared to leading 8B models. Reka Edge is a 7B VLM built for latency-sensitive apps: real-time video analysis, agentic workflows, on-device deployment reka.ai/news/reka-edge…

Many VLMs look impressive in evals. But when you try to deploy them, the tradeoffs show up fast: memory, latency, token cost, throughput. We built Reka Edge to change that: frontier visual intelligence that's actually deployable. And it doesn't sacrifice quality for speed. Best-in-class for its size across: • Video & Image Understanding (VQA-V2, MLVU, MMVU) • Object Detection & Grounding (RefCOCO) • Tool Use (Mobile Actions) • Reliability & Truthfulness (VideoHallucer) It even approaches Gemini 3 Pro at a fraction of the cost!

@robodadg 🔥🔥🔥

@RekaAILabs Approaching Gemini 3 Pro at 7B params. Big if true.

Meet Reka Edge – Our next-generation vision language model for physical AI. Uses 3x fewer input tokens and achieves 65% faster throughput compared to leading 8B models. Image understanding, video analysis, object detection, and tool use. Built for Action. Fast enough for production, deployable anywhere. Read more: reka.ai/news/reka-edge…

New Reka community node for n8n is live 🚀 AI powered video clipping + image analysis inside your workflows. Create short clips. Analyze visuals. Ask questions about images. No code required. 🔗 reka.ai/news/introduci… #AI #n8n #automation

New n8n template 🚀 Automatically generate AI clips from YouTube videos, triggered on every new upload. Perfect for repurposing long-form content into social-ready shorts. Details & template here: dev.to/reka/automatic…

Smart cameras shouldn’t just detect motion. They should understand events. Reka Vision reasons across video, audio, and time,  reducing false alerts and delivering contextual, human-level insights. We've created a new standard for vision AI 👁️ 🔗 reka.ai/news/reka-visi…

"It's no longer about creating more content — it's about unlocking the value in the content you already own." At #NABShowNY, our own @meenal_n shared how we're helping broadcasters transform decades of footage into new revenue through: 🎬 Discovering moments that matter within massive archives 💰 Unlocking new ad inventory with context-aware premium placements 📊 Quantifying every brand appearance to prove ROI 🔍 Searching video archives using natural language ✂️ Clipping and repurposing content to share across multiple platforms 🧠 Conducting agentic research on audience sentiment, creator performance, and narrative trends — automatically Local media still holds one advantage no AI can replicate: trust. We're excited to help amplify that trust in the next era of AI-driven storytelling. See full video here: youtu.be/EQIbWsc6egE

Unlock the full potential of your AI chat applications with Reka's powerful and completely free MCP Server! This video guides you through integrating online search capabilities, AI-powered fact-checking, and more directly into your VS Code projects. GitHub: link.reka.ai/research-mcp

Meet us at #SFTechWeek! We're excited to be hosting an evening of hands-on building and networking on Tuesday, October 7, during SF Tech Week! Learn how to use agentic web search and Vision AI in your solution. This will be a chance for you to have your hands-on with Reka's latest tools to power real-world workflows, followed by a happy hour mixer! Connect with fellow developers, founders, and partners over drinks and bites, and still make it to the rest of your Tech Week plans! 📅 Tuesday, October 7 🕓 4:00pm - 7:00pm 📍 FIDI, 330 Pine St, San Francisco, CA 94104 🔗 We have a few spaces left - Register here: link.reka.ai/sftechweek-ll

Introducing Parallel Thinking for Reka Research! Instead of one line of reasoning, we explore multiple paths in parallel, then resolve the best answer. Big accuracy gains on Research-Eval (+4.2) and SimpleQA (+3.5). Now live in the Reka Research API!

@notnotrishi Thanks! Our setup is different (the model retrieves information by browsing the web or internal files, so hallucinations are rare), but it might still be worth giving it a try.

@RekaAILabs this is simple but useful! i wonder if applying SLED to refine each candidate answer generated by parallel thinking makes it even better arxiv.org/pdf/2411.02433

Parallel thinking is now available in our API. Choose your level of reasoning power: • Low mode: $35 / 1k reqs • High mode: $60 / 1k reqs Also available in our free API tier! Learn more: reka.ai/news/introduci…

How it works: 1️⃣ Parallel generation: Reka Research generates N candidate answers concurrently. Speed stays nearly the same! 2️⃣ Resolution: A separate model, trained on verifiable problems, resolves among these candidates to produce the final answer. Simple idea → big impact!

Tech Week is taking over SF on Oct 7! We are hosting Dev Workshop & Mixer, where you'll learn something new and network with fellow devs and founders. Join us here: link.reka.ai/sftechweek #SFTechWeek