Releases·Model Releasefeatured
NVIDIA's Nemotron-Labs-3-Competitive-Coding 550B model outscores top human at IOI 2026
NVIDIA released Nemotron-Labs-3-Competitive-Coding-550B-A55B-NVFP4, a competitive-programming specialist fine-tuned from Nemotron-3-Ultra-550B-A55B. Combined with its GenCorrect test-time compute strategy, it scored 535.4/600 on the IOI 2026 problem set under official contest conditions — above the gold threshold and the top human contestant's 498.27 — reportedly the first AI to outscore the best human on an IOI set. Weights are open for commercial and non-commercial use.
273 engagement1 sourceMon, Sep 28, 2026, 11:45 PM
AI Overview
- NVIDIA's Nemotron-Labs-3-Competitive-Coding-550B-A55B-NVFP4, fine-tuned from Nemotron-3-Ultra, scored 535.4/600 at IOI 2026, beating the top human's 498.27 and gold threshold 361.12
- The model was fine-tuned one epoch on 477,642 synthetic reasoning traces distilled from GLM-5.2 across 22,000 problems in 16 contest families; GLM-5.2 chosen over a DeepSeek-V4-Flash teacher for ~30% shorter generations
- Combined with GenCorrect, an iterative test-time compute strategy generating diverse candidates and refining them on evaluator feedback, it was evaluated live under official IOI 2026 constraints — the first AI reported to outscore the top human contestant on an IOI problem set
Entities
Related
Releases·Model Release
One Model Family, Two Gold-Level Results: Fine-Tuning Nemotron for IOI and IMO
NVIDIA reports that Nemotron 3, specialized with SFT, RL, and generate-verify-refine inference, reached gold-medal level at IMO 2026 (30/42) and IOI 2026 (535.4/600). The company released the checkpoints, training datasets, and inference pipelines on Hugging Face and GitHub to make the recipe reproducible.
NVIDIA reports that Nemotron 3, specialized with SFT, RL, and generate-verify-refine inference, reached gold-medal level at IMO 2026 (30/42) and IOI 2026 (535.4/600). The company released the checkpoints, training datasets, and inference pipelines on Hugging Face and GitHub to make the recipe reproducible.
0 engagement1 sourcerssWed, Oct 7, 2026, 12:45 PM
Social·Community
NVIDIA bets Vera Rubin GPUs on NVFP4, its FP4 precision for LLM training and inference
A YouTube video by bycloud explains NVFP4, the 4-bit floating-point precision NVIDIA is building its upcoming Vera Rubin GPUs around, and the training tricks that make FP4 stable for LLM training. The video draws on NVIDIA's NVFP4 pretraining paper, its Quantization-Aware Distillation report, and Nemotron 3 technical documents. It matters to practitioners because low-precision training is central to NVIDIA's newest GPU designs.
A YouTube video by bycloud explains NVFP4, the 4-bit floating-point precision NVIDIA is building its upcoming Vera Rubin GPUs around, and the training tricks that make FP4 stable for LLM training. The video draws on NVIDIA's NVFP4 pretraining paper, its Quantization-Aware Distillation report, and Nemotron 3 technical documents. It matters to practitioners because low-precision training is central to NVIDIA's newest GPU designs.
1.3k engagement1 sourceyoutubeMon, Sep 21, 2026, 04:23 PM
Releases·Model Release
llama.cpp PR adds support for NVIDIA Nemotron-3-Puzzle-75B-A9B, a 75.3B-parameter hybrid Mamba-MoE model
A llama.cpp pull request by YanissAmz adds support for NVIDIA's Nemotron-3-Puzzle-75B-A9B, a pruned-down hybrid Mamba-MoE-Attention model. The pruned variant shrinks the parent 120.7B-total-parameter model to 75.3B total / 9.3B active parameters, making it locally runnable today, though Multi-Token Prediction is not yet supported in llama.cpp.
A llama.cpp pull request by YanissAmz adds support for NVIDIA's Nemotron-3-Puzzle-75B-A9B, a pruned-down hybrid Mamba-MoE-Attention model. The pruned variant shrinks the parent 120.7B-total-parameter model to 75.3B total / 9.3B active parameters, making it locally runnable today, though Multi-Token Prediction is not yet supported in llama.cpp.
37 engagement1 sourceredditThu, Sep 3, 2026, 07:35 AM
Releases·Model Release
NVIDIA releases NemotronLabs VoiceChat-11B full-duplex model on Hugging Face
NVIDIA has released the NemotronLabs VoiceChat-11B model on Hugging Face, featuring full-duplex voice interaction. The model enables real-time, bidirectional conversation, allowing users to speak and listen simultaneously. This release advances voice AI capabilities for developers building conversational interfaces.
NVIDIA has released the NemotronLabs VoiceChat-11B model on Hugging Face, featuring full-duplex voice interaction. The model enables real-time, bidirectional conversation, allowing users to speak and listen simultaneously. This release advances voice AI capabilities for developers building conversational interfaces.
249 engagement1 sourceredditMon, Aug 3, 2026, 10:24 PM
Releases·Model Releasefeatured
NVIDIA releases Nemotron 3.5 Lightning 30B A3B BF16 on Hugging Face
NVIDIA has released Nemotron 3.5 Lightning, a 30B-parameter mixture-of-experts model with 3B active parameters, designed for fast and efficient long-running agentic AI workloads. The release includes BF16 and NVFP4 quantized versions on Hugging Face, positioning it as the highest-efficiency model in its class for specialized tasks within multi-agent systems.
NVIDIA has released Nemotron 3.5 Lightning, a 30B-parameter mixture-of-experts model with 3B active parameters, designed for fast and efficient long-running agentic AI workloads. The release includes BF16 and NVFP4 quantized versions on Hugging Face, positioning it as the highest-efficiency model in its class for specialized tasks within multi-agent systems.
2301.1k engagement5 sourcesreddit, youtube, rss +1Tue, Aug 11, 2026, 01:19 PM