Crusoe for Developers

Before the PyTorch Conference, Crusoe and Cerebral Valley are bringing builders to Shack15 to push past the one-shot agent: multi-agent teams that plan, delegate, and recover from failure on their own. Most agent demos still need a human in the loop. We want to see what happens when you take that human out. Build on Crusoe Managed Inference with leading open models to win $25,000 in cash prizes + an NVIDIA RTX 5090. 📅 Sat, Oct 17 · 9:00 AM – 10:00 PM PDT 📍 Shack15, San Francisco 👥 Teams of up to 4 This one's fully in-person, and spots are limited — all attendees must apply and be approved. Register here: https://lnkd.in/g4UNtmCE Build a team of agents you can actually walk away from. See you there. #AI #hackathon

24 hours. 350 builders, product developers, and engineers. What can you build in a single day? Today, on Day Zero of The AI Conference, Crusoe is sponsoring an all-day hackathon in SF. The challenge behind Hack Day is simple: take an idea from zero to a working product in just one day. The winning team will take the conference's main stage tomorrow to pitch a panel of top venture capitalists. We’ll be here all day. Come meet the Crusoe for Developers team and let’s build. Space is limited. Register: 🔗 https://lnkd.in/ga6Ei9TQ

Our next Crusoe AI Talk is coming to San Francisco on 𝗢𝗰𝘁𝗼𝗯𝗲𝗿 𝟳. Erwan Menard, our SVP of Product Management, joins Jeffrey Morgan (co-founder, Ollama) and Michael Elabd (co-founder, Trajectory) for a panel on one of the biggest decisions AI teams face: 𝘞𝘩𝘦𝘯 𝘴𝘩𝘰𝘶𝘭𝘥 𝘺𝘰𝘶 𝘶𝘴𝘦 𝘢 𝘤𝘭𝘰𝘴𝘦𝘥 𝘧𝘳𝘰𝘯𝘵𝘪𝘦𝘳 𝘮𝘰𝘥𝘦𝘭, 𝘢𝘯 𝘰𝘱𝘦𝘯 𝘮𝘰𝘥𝘦𝘭, 𝘰𝘳 𝘵𝘳𝘢𝘪𝘯 𝘺𝘰𝘶𝘳 𝘰𝘸𝘯? They’ll dig into how their teams think about cost, control, performance, and speed when making that decision. 𝗪𝗲𝗱𝗻𝗲𝘀𝗱𝗮𝘆, 𝗢𝗰𝘁𝗼𝗯𝗲𝗿 𝟳 | 𝟱:𝟬𝟬–𝟴:𝟯𝟬 𝗣𝗠 | 𝗦𝗮𝗻 𝗙𝗿𝗮𝗻𝗰𝗶𝘀𝗰𝗼 Space is limited. 𝗥𝗲𝗴𝗶𝘀𝘁𝗲𝗿 𝗵𝗲𝗿𝗲: https://luma.com/lnnlfa87

The AI agent ecosystem is moving fast, and we're excited to be building in it. At the DevNetwork AI/ML Hackathon 2026, the winning team built SkinMatch AI, an app that analyzes your skin from a photo and generates a personalized skincare routine. Under the hood: 1. Nvidia Nemotron running on Crusoe Cloud Managed Inference for routine generation 2. Llama 3 and Qwen on Crusoe as fallback models 3. TrueFoundry AI Gateway routing between them so the app stays up even when a model goes down That's a production-grade resiliency pattern, built in a weekend. Watch the full demo: https://lnkd.in/d9vfJWih And it sparked something bigger. Crusoe is now an official integration with TrueFoundry, so any team can route to Crusoe Managed Inference through TrueFoundry's gateway with the same fallback and observability the hackathon winner used. Developers: what would you build with Nemotron on Crusoe? Get started → https://lnkd.in/d899VdsC

Crusoe will be at AI Infra Summit, Sep 15–17, Santa Clara. Come talk to our team about our vertically integrated infrastructure, sustainable energy for AI, managed AI services, open models, serverless fine-tuning, and more.

GLM 5.3 and GLM 5.3 Flash are now in Crusoe Intelligence Foundry. ➡️ GLM-5.3: Same base as GLM-5.2, with every gain coming from scaled post-training. Terminal-Bench 3.0 moved from 4.6 to 28.3. On Z.ai's in-house Code Bench it hits 34.5% at max reasoning effort using roughly 75K output tokens per task, where GLM-5.2 spent 96K to reach 23.4%. ➡️ GLM-5.3-Flash: MIT licensed with weights already released. Roughly 9x cheaper for workloads that do not need the frontier tier. Try both models today without provisioning a single cluster. → https://lnkd.in/gMVD8Q7w

𝗥𝘂𝗻𝗻𝗶𝗻𝗴 𝗢𝗺𝗮𝗿𝗰𝗵𝘆? 𝗬𝗼𝘂𝗿 𝗰𝗼𝗱𝗶𝗻𝗴 𝗮𝗴𝗲𝗻𝘁 𝗰𝗮𝗻 𝗻𝗼𝘄 𝗿𝘂𝗻 𝗼𝗻 Crusoe. Omarchy ships OpenCode as its default coding agent. OpenCode pulls its provider catalog from models.dev. And Crusoe is now in models.dev. That means every Omarchy dev already has a path to open-weight models on Crusoe Managed Inference, with nothing to install: 1️⃣ Open a terminal and type c to launch OpenCode 2️⃣ Run /connect, pick Crusoe, and paste your API key from console.crusoecloud.com 3️⃣ Pick a model with /models and start working Or just export CRUSOE_API_KEY and OpenCode picks it up. What's on the menu: DeepSeek, Qwen3, Kimi K2, GLM, Llama, Nemotron, and gpt-oss, all served through our OpenAI-compatible endpoint at https://lnkd.in/g8_mUUD2. Same agent, same keybindings, same Omarchy theme. Different silicon underneath. This is the second coding-agent surface Crusoe has landed in recent weeks. We're working down the list. Docs: https://lnkd.in/gEqUPXEZ Provider listing: https://lnkd.in/g6wTanMg #Omarchy #OpenCode #Crusoe #AIInfrastructure #OpenSource #DeveloperTools #LLM

New guide from Crusoe: 𝘖𝘱𝘦𝘯, 𝘊𝘭𝘰𝘴𝘦𝘥, 𝘰𝘳 𝘉𝘰𝘵𝘩. 𝘈 𝘧𝘳𝘢𝘮𝘦𝘸𝘰𝘳𝘬 𝘧𝘰𝘳 𝘳𝘶𝘯𝘯𝘪𝘯𝘨 𝘈𝘐 𝘸𝘰𝘳𝘬𝘭𝘰𝘢𝘥𝘴. Where you run each workload sets your cost floor, your latency, and how much control you keep. This guide gives developers a practical framework for making that call. Inside: 1. Five questions that decide where a workload runs: volume, latency, customization, data, and cost at scale. 2. Where closed frontier models win and where fine-tuned open models take over. 3. A five-property test to verify any platform's "your model" claims before you sign. 4. What production fidelity actually takes, because holding weights and serving them well are different capabilities. Read the full guide: https://lnkd.in/g7DAkgYz

The Crusoe for Developers team is heading back to Paris. Connor Guerrero, our Senior Developer Relations Manager at Crusoe, will take the stage at AI Engineer Paris, the second edition of the conference organized by Mistral, happening September 23-24 at STATION F. His talk covers GPU failures at scale: what they cost, and the patterns that make them a non-event for your training and inference workloads. The two-day event brings together the global community of AI engineers for inspiring talks, hands-on workshops, 6 dedicated tracks, and an Expo Hall featuring some of the best minds in AI. Grab your Super Early Bird ticket with code SPEAKER-EARLYBIRD and come say hi to the Crusoe team at STATION F.