NVIDIA's sub-1B Nemotron 3.5 Parse tops ParseBench as best open-weight doc parser
llama_index · x · 2026-10-22
- Jerry Liu (LlamaIndex) congratulates NVIDIA on Nemotron 3.5 Parse, which takes the #1 spot on ParseBench — a real-world document understanding benchmark covering 150+ frontier, open-weight, and specialized models — while staying under 1B parameters for blazing-fast inference; Jensen Huang mentioned it at GTC Berlin.
- It ships alongside the expanded Nemotron Retriever collection: Nemotron 3.5 Embed 8B/1B (BF16 and NVFP4) and Nemotron 3.5 Rerank 8B, all openly and commercially licensed, bringing multimodal retrieval, relevance ranking, and document extraction to enterprise agent workflows and cutting tokens wasted on poor retrieval.
Related event: NVIDIA Open-Sources Nemotron 3.5 Retrieval Model Family(3 posts)→
More from coding & agent
- TraceForge synthesizes reasoning traces from answers, beating Claude's own CoT by +3.7/+7.1 — IndependentResearcher121 · 2026-10-22
- Internet Archive fends off PHP attacks overnight, using Claude to patch old code — kingslyj · 2026-10-22
- Combining Ghidra, AI Reverse Engineering and WebGPU Sparks an "Absolute Renaissance" — thursdai_pod · 2026-10-22
- As AI writes more code, observability becomes the key to catching agent business-logic failures — mipsytipsy · 2026-10-22
- Reordering Identical C Struct Fields Cuts Size From 24 to 16 Bytes — cneuralnetwork · 2026-10-22
- fal employee shows Claude + fal generating a game in 60 minutes — OdinLovis · 2026-10-22