{"id":"smaug-nemotron-3-5","slug":"smaug-nemotron-3-5","name":"Smaug-Nemotron-3.5","orgId":"abacus-ai","family":"smaug-nemotron-3-5","releaseDate":null,"availability":"announced","openWeights":true,"license":null,"paramsTotal":null,"paramsSource":null,"contextWindow":null,"knowledgeCutoff":null,"modelCardUrl":"https://x.com/bindureddy/status/2113480704764076310","hfRepo":null,"modalities":["text","image","video"],"aliases":["Smaug-Nemotron-3.5","smaug-nemotron-3.5","smaug nemotron 3.5","smaug-nemotron","abacusai/Smaug-Nemotron-3.5","abacusai/Smaug-Nemotron-3.5-NVFP4"],"aliasNote":null,"origin":"tracked","tracked":{"trackedSince":"2026-10-23","announced":"2026-10-23","blurb":"Abacus.AI's agentic-coding fine-tune of NVIDIA's Nemotron 3.5 Super VL (120B total / 12B active), supervised-fine-tuned on multi-turn agentic coding trajectories. Open weights on Hugging Face in BF16 and NVFP4 under the OpenMDW-1.1 license inherited from the base; Abacus.AI says it runs on a single DGX Spark, H200 or GB200.","api":"none","openWeights":"yes","params":{"text":"120B total / 12B active","source":"Hugging Face model card (abacusai/Smaug-Nemotron-3.5)"},"facts":[{"label":"Base model","value":"nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16"},{"label":"Training","value":"Supervised fine-tuning on agentic multi-turn coding trajectories"},{"label":"Architecture","value":"Hybrid Mamba-2 + mixture-of-experts + attention, with a C-RADIO vision encoder and an MTP head (from the base)"},{"label":"Input / output","value":"Text, images and video in; text out. Reasoning can be switched on or off"},{"label":"Context","value":"Up to 1M tokens"},{"label":"Weights","value":"Hugging Face: abacusai/Smaug-Nemotron-3.5 (BF16) and abacusai/Smaug-Nemotron-3.5-NVFP4; not gated"},{"label":"License","value":"OpenMDW-1.1 (inherited from the NVIDIA base)"},{"label":"Hardware","value":"Lab says one DGX Spark, one H200 or one GB200 is enough"}],"claims":["Lab claim (self-reported, model card, Smaug vs the base): SWE-bench Verified 72.3 vs 71.1; SWE-bench Multilingual 68.9 vs 67.7; Terminal-Bench 2.1 69.1 vs 60.7; DeepSWE 27.4 vs 24.7; NL2RepoBench 32.2 vs 30.2; LiveBench 2.0 Agentic 18.1 vs 15.8.","Lab claim: the best US open-source agentic model that can run on a single machine (Abacus.AI CEO's announcement)."],"sources":[{"kind":"lab","title":"Bindu Reddy (Abacus.AI CEO) announcing Smaug-Nemotron-3.5 on X","url":"https://x.com/bindureddy/status/2113480704764076310","date":"2026-10-23"},{"kind":"repo","title":"abacusai/Smaug-Nemotron-3.5 on Hugging Face","url":"https://huggingface.co/abacusai/Smaug-Nemotron-3.5","date":"2026-10-23"},{"kind":"repo","title":"abacusai/Smaug-Nemotron-3.5-NVFP4 on Hugging Face","url":"https://huggingface.co/abacusai/Smaug-Nemotron-3.5-NVFP4","date":"2026-10-22"}],"firstBenchmark":null},"organization":{"id":"abacus-ai","name":"Abacus.AI","slug":"abacus-ai","country":null,"url":null,"color":"#11110F","mark":"A"},"price":null,"answer":{"word":"Pending benchmarks.","scope":"Announced Oct 23, 2026; no independent results yet.","lead":"As of Nov 4, 2026, Smaug-Nemotron-3.5 is pending benchmarks on isitsota: announced Oct 23, 2026, with no independent results on the boards we track yet.","confidence":"low"},"status":"PENDING_BENCHMARKS","verdicts":[]}