gordana neskovic @nesgor

Today, we're announcing Baseten for Model Labs. We believe the AI landscape will be made up of a diverse ecosystem of specialized, closed-weight models running alongside open-weight ones. Baseten for Model Labs is the platform built to enable labs to monetize, distribute, and scale their closed models on Baseten infrastructure.

GPU per hour is just the tip of the inference iceberg. 🧊 Agents can use 10 to 100x more tokens per task than a single chat exchange. @NVIDIA Nemotron 3 Ultra, NVIDIA Dynamo, and fastokens are how we're building for what's below the surface: goodput and throughput. 🔗 crusoe.ai/resources/blog… #AgenticAI #Nemotron #AIInfrastructure

Your structured output fills with endless whitespace instead of JSON. Engine bug? Probably not. Constrained decoding masks invalid tokens, but it can't pick which valid token comes next. Whitespace is valid JSON grammar. If your prompt doesn't signal the expected fields, the model can stall on newlines until the token limit. 🧑‍🔧The fix: put a concrete example output in your prompt. Bonus: on FriendliAI Endpoints, response_format is up to 21% faster — narrowing the token set boosts draft acceptance in speculative decoding. Prompt guidance + constrained decoding. You need both. 🔗 friendli.ai/blog/structure…

I'm very excited to share that @nvidia just released Nemotron-3-Embed: two multilingual embedding models for retrieval The 8B takes the top spot on RTEB. - Nemotron-3-Embed-1B-BF16 (2048-dim) - Nemotron-3-Embed-8B-BF16 (4096-dim) Both OpenMDW-1.1, ready for commercial use. 🧵

Today we released Nemotron 3 Embed 8B and it reached #1 overall on RTEB 🏆 RTEB benchmarks retrieval accuracy across real-world tasks. Better retrieval gives agents more relevant context, helping improve response accuracy.

Retrieval is where AI agents live or die. Miss the right context and the agent reasons from the wrong place. @nvidia Nemotron 3 Embed is now on DeepInfra with Day-0 support, - the leading, embedding model for enterprise search. RAG, and agentic retrieval. Tops the RTEB leaderboard. Two sizes, live now 👇

🚀Day 0 Availability: NVIDIA Nemotron 3 Embed (8B & 1B) is live on @Bitdeer_AI Model Studio. Frontier retrieval accuracy for RAG, enterprise search, code search & agentic workflows with the 1B model retaining 95%+ of 8B accuracy for high-volume production. Open weights, open recipes. Build now 👉 bitdeer.ai/en/model/explo… #Nemotron #RAG #AgenticAI

whoooo hooooo! @NVIDIAAI is COOKING! I have been very excited about this and waiting to break the news! This is going to get incorporated into my Praxis agent ASAP! NVIDIA Nemotron 3 Embed Ranks #1 Overall on RTEB, Advancing Agentic Retrieval Retrieval is critical in multi-step agentic workflows where poor retrieval can cause agents to fetch irrelevant context, re-query, waste token budget, and carry noise into later reasoning steps. Today, we are releasing NVIDIA Nemotron 3 Embed, a collection of open and commercially available embedding models designed to improve retrieval quality while giving developers practical deployment options for production-scale RAG, agentic retrieval, code retrieval, and agent memory. The collection includes three open models that achieve state of the art retrieval across the accuracy-efficiency curve, led by an 8B model that tops the RTEB leaderboard and efficient 1B variants built for production-scale deployment: ----------------------------------------------------- Nemotron-3-Embed-8B- BF16 Flagship Quality Anchor: The flagship embedding model, ranking #1 on RTEB. Precision-critical retrieval and high-stakes enterprise RAG huggingface.co/nvidia/Nemotro… ---------------------------------------------------- Nemotron-3-Embed-1B-BF16 High-Efficiency Standard: A high-efficiency model for productionretrieval where latency and costmatter. Cost- and latency-sensitive production serving huggingface.co/nvidia/Nemotro… ---------------------------------------------------- Nemotron-3-Embed-1B-NVFP4 Hardware-Accelerated Variant: A Blackwell-optimized variant for high-throughput retrieval with a smaller memory footprint. Ultra-high-throughput and massive-scale infrastructure huggingface.co/nvidia/Nemotro…

NVIDIA just released Nemotron 3 Embed. These are the models you pair with an LLM to find the right context for RAG, agent memory, code search, etc. The 8B model is now #1 on the RTEB multilingual leaderboard. NVIDIA also released 1B BF16 and NVFP4 versions for teams that need a smaller deployment footprint. The weights and training recipes are open too. Try the 1B model through NVIDIA NIM build.nvidia.com/nvidia/nemotro… huggingface.co/collections/nv…

🚦 The zero-shot model misses a visible traffic signal. After post-training Cosmos 3 Nano with LoRA, it correctly identifies the intersection, and overall WTS validation accuracy increases from 54.41% to 87.14%. 📈 With NVIDIA TAO AutoML, accuracy reaches 93.35%. All in under a day.

keyboard_arrow_left Previous keyboard_arrow_right Next

Excited to see this out in the open. We've been working with @LangChain to tune their Deep Agents harness for Nemotron 3 Ultra. The result is a fully open, customizable agent that delivers frontier performance at a fraction of the cost of leading closed models.

LangChain @LangChain

4 weeks ago

We tuned the harness for @NVIDIAAI Nemotron 3 Ultra. Benchmark-leading performance. 10x lower inference costs. ✅ An aggregate score of 0.86 at a cost of $4.48 ✅ The closest-performing model: $43.48 langchain.com/blog/langchain…

16 24 188 57K 56

The Nemotron family just passed 100M downloads! Huge thank you to the community building with us and showing what’s possible with open models. Cheers to OSS 🍾

Nemotron 3 Ultra is fast and genuinely good Compared it with 3 frontier models: DeepSeek V4, MiniMax M3, and Qwen 3.7 Max on 2 prompts very impressive results

📢 EnterpriseOps-Gym is now accepted to ICML 2026 🇰🇷✨ website: enterpriseops-gym.github.io 🧩 1,150 expert-curated tasks 🏢 8 enterprise domains 🧰 512 tools ✅ Deterministic verifiers (Outcome + Integrity + Compliance) 📦 Fully containerized, no enterprise instance required 📊 𝗙𝗿𝗲𝘀𝗵 𝘂𝗽𝗱𝗮𝘁𝗲: GPT-5.5 numbers are out (alongside the strongest open and closed baselines). We will keep updating as new models drop because long-horizon reliability is moving fast, and we want to stay current. One exciting (and humbling) signal: we’ve already seen frontier lab teams experimenting with EnterpriseOps-Gym to stress-test and improve their agents: including folks at OpenAI, Mistral AI, and NVIDIA AI. 🙏 📈 𝗘𝗮𝗿𝗹𝘆 𝗿𝗲𝘀𝘂𝗹𝘁𝘀 𝗮𝗿𝗲 𝗽𝗿𝗼𝗺𝗶𝘀𝗶𝗻𝗴: Top open models  including NVIDIA Nemotron Super  are showing strong performance, in some cases competing with frontier models. @shiva_malay @sagardavasam @PShravannayak @turingcom @jonsid @ServiceNowRSRCH @Mila_Quebec

.@nvidia Nemotron 3 Nano Omni. First on DigitalOcean. 🤖🆕 One model. Text, vision, & native tool use. Document intelligence, multimodal RAG, agentic workflows. One API call, one bill. Generally available today on DigitalOcean Serverless Inference. Try it in the Multimodal Playground in the comments. ⬇️

Nemotron 3 Nano Omni is available locally on Ollama! This requires the latest Ollama 0.22 release.

NVIDIA AI @NVIDIAAI

3 months ago

Meet Nemotron 3 Nano Omni 👋 Our latest addition to the Nemotron family is the highest efficiency, open multimodal model with leading accuracy. 30B parameters. 256K context length. 🧵👇

93 186 1K 460K 501

NVIDIA Nemotron 3 Nano Omni is live on Nebius Token Factory. A major step forward for multimodal reasoning, bringing video, audio, images, documents, and text into a single model. Bringing these workloads into production requires consistent performance, low latency, and efficient scaling. Nemotron 3 Nano enables this with a unified efficient architecture, while we focus on the infrastructure layer that ensures it performs reliably under real-world load. From capability to production. Try it today 👇

Congrats to @nvidia on releasing NVIDIA Nemotron 3 Nano Omni! GMI is supporting it on Day 0. Nemotron 3 Nano Omni is an open multimodal model with highest efficiency that powers sub-agents to complete tasks faster across vision, audio, and language. · 30B-A3B MoE · 256K context · ~2x throughput · ~2.5x less compute on video start on GMI today 👇 gmicloud.ai/en/blog/runnin…

NVIDIA Nemotron™ 3 Nano Omni is deployed on Vultr. Available on dedicated GPU clusters or serverless inference accelerated by @nvidia Dynamo 1.0, Nemotron™ 3 Nano Omni accelerates faster, more efficient, more accurate enterprise agent task completion: blogs.vultr.com/nvidia-nemotro…

NVIDIA Nemotron™ 3 Nano Omni is live on OpenRouter. An open 30B-A3B multimodal model for agentic workflows: text, image, video, and audio in → text out, with a 256k context window and efficient MoE architecture for computer use, documents, and AV reasoning.