Decompute AI

We are at Confidential Computing Summit 2026. Want to learn about our heterogeneous agentive compute fabric Echelon? Drop in!

We guarantee your Claude costs drop. If they don’t, your trial gets extended. Decompute Gateway: from cutting tokens → owning your AI. One-click VS Code install. Zero data sent to us. Runs local. claude.decompute.run

Cut Claude Code tokens before they hit your bill for free 💵 84,000 raw tokens ↓ Decompute Claude Gateway -78% on this request ↓ 18,700 sent tokens Same Claude Code. Same workflow. Fewer tokens. 🧮 claude.decompute.run

We’re sharing Nebula-S-SVMS2-3B-Internal, an internal evaluation release from Decompute. Internal evals for this release report: GPQA: 86.85 HMMT Nov 2025: 80.00 GSM8K: 92.00 MMLU-Pro: 83.00 Model: lnkd.in/gterpHN4

We gave the same messy operating task to an intern and to Nebula-S: customer interviews + support tickets + partial docs + conflicting notes →produce a decision memo The intern took hours. Nebula-S-4B produced a strong first pass in minutes. ~15× faster time-to-first-draft.

Introducing BlackBird Enterprise — a structurally private AI platform that runs where your data already lives: on-device, on-prem, or in your cloud. Built for orgs that refuse to compromise on: • Data residency, compliance & guardrails • Structural privacy by design • On-device/on-prem RAG, document AI, copilots • Distributed inference & training (DGX, multi-tenant) • Works with files, folders, KBs, Excel, SharePoint, GDrive, Confluence, Notion • Voice → transcription, summaries, Q&A • Powered by our Nebula-S models beating Qwen & DeepSeek on MMLU/ARC/GSM8K If you’re navigating AI + privacy + cost, this is for you 👇 lnkd.in/gpQViNkn #AI #EnterpriseAI #PrivacyByDesign #OnDeviceAI #GenAI

🔥 New Release on high demand: Now train Open AI’s GPT-OSS 20B for personalized agents locally on Mac with BlackBird on demand: ⚡ On-device training 🔒 Full privacy 🛠️ No cloud, no compromise decompute.run/releases (We received several messages from everyone requesting this release, and it seems MLX was completely broken and did not support fine-tuning, so we fixed it ourselves.) #BlackBird #Decompute #OpenAI #oss

🚀 Introducing Kestrel VLM: our vision-language model that self-improves after deployment and leaps 30 % past DeepMinds' SigLIP-2, unlocking a new “fusion-space” frontier for enterprise vision AI. Kestrel can learn from just 5-10 examples, unlocking reinforcement learning at the edge for vision models—no cloud retraining needed. Adapt on the fly for real-world understanding, OCR, charts, and more. Drive accuracy higher, boost privacy, and personalize models directly on your device. Ready for chat-grade speed and next-level autonomy? Try Kestrel VLM now. decompute.run/post/introduci… #VisionAI #GenAI #KestrelVLM #AIatTheEdge

🚨 Calling all builders in SF! $50K up for grabs! 🚨 We’re thrilled to sponsor the A2A Agents Hackathon in SF! 💥 Build on-device AI agents with the Blackbird SDK 🧠 Multimodal, fast, secure, insanely powerful 🏆 Judged by Decompute’s own @hinadixit 🎯 Sign up: lu.ma/aoc5opn4 #AIAgents #Hackathon #OnDeviceAI #Decompute #BlackbirdSDK

📢 Exciting news! The BlackBird-SDK is now open source! 🎉 - Local AI inference | - Real-time streaming | - Secure file processing 👉 Contribute: github.com/Decompute-AI/Blac… #OpenSource #AI #MachineLearning

Finally: AI that doesn't spy on your users. Build it with BlackBird 👇 **BlackBird SDK Launched!** Build private, on-device AI agents in minutes. No cloud, no compromises. **Why it’s game-changing:** 🔒 **100% Private**: Data never leaves the device ⚡️ **<100ms latency**: Zero API hops 🛠️ **Any open model**: No vendor lock-in **For developers:** - Electron/React Native-ready - Fine-tune locally (no cloud GPUs) - Audio/vision/function-calling in one API 🚀 **Perfect for**: Finance | Healthcare | Legal | Offline apps 🔗 Docs: lnkd.in/gTv8SyUT 🧑‍💻 Contribute: lnkd.in/geTrM5Gv #OnDeviceAI #PrivacyFirst #EdgeComputing

Confused when to use a single dense model, multiple SLMs and when to use SMoEs? A dense 6B model is like a generalist employee who “knows a bit of everything.” But when you need legal precision, you want a lawyer, not someone who also handles IT tickets and HR disputes. That’s what MoE gives you - specialized, composable intelligence for different domains, where each Expert is a master and not a generalist. Then why not use multiple SLMs? Because if you need a single user prompt to juggle multiple domains in real time → use an MoE. But if you have independent workflows, tolerant of a 1-second model swap or running in parallel services? → deploy several SLMs. Now, with Decompute you can build your own MoEs, you can also hand pick which experts you want to fine-tune with our SDK. Only 5-15 % of experts wake up, slashing storage by 90 % and fine-tune time by 30 % . That’s 75-95 % fewer trainable parameters vs. full fine-tunes, lowering the cost to train by 50%!