litlas
Log inCreate account

Olmo 3

Authors: Team OLMo, :, Allyson Ettinger, Amanda Bertsch, Bailey Kuehl, David J. Graham, David Heineman, Dirk Groeneveld, Faeze Brahman, Finbarr Timbers, Hamish Ivison, Jacob Morrison, Jake Poznanski, Kyle Lo, Luca Soldaini, Matt Jordan, Mayee Chen, Michael Noukhovitch, Nathan Lambert, P N Walsh, Pradeep Dasigi, Robert Berry, Saumya Malik, Saurabh A. Shah, Scott Geng, Shane Arora, Shashank Gupta, Anderson, Taira, Xiao, Teng, Tyler Murray and 39 more - arXiv 2025 cited by 277

We introduce Olmo 3, a family of state-of-the-art, fully-open language models at the 7B and 32B parameter scales. Olmo 3 model construction targets long-context reasoning, function calling, coding, instruction following, general chat, and knowledge recall. This release includes the entire model flow, i.e., the full lifecycle of the family of models, including every stage, checkpoint, data point, and dependency used to build it. Our flagship model, Olmo 3 Think 32B, is the strongest fully-open thinking model released to-date.

The Llama 3 Herd ofModelsThe Llama 3 Herd of ModelsDeepSeek-V3 TechnicalReportDeepSeek-V3 Technical ReportDataComp-LM: In searchof the next generation…DataComp-LM: In search of the next generation of training sets for language modelsMMLU-Pro: A More Robustand Challenging…MMLU-Pro: A More Robust and Challenging Multi-Task Language Understanding BenchmarkK2-V2: A 360-Open,Reasoning-Enhanced LLMK2-V2: A 360-Open, Reasoning-Enhanced LLMNVIDIA Nemotron Nano 2:An Accurate and…NVIDIA Nemotron Nano 2: An Accurate and Efficient Hybrid Mamba-Transformer Reasoning ModelOpenThoughts: DataRecipes for Reasoning…OpenThoughts: Data Recipes for Reasoning ModelsMiMo: Unlocking theReasoning Potential of…MiMo: Unlocking the Reasoning Potential of Language Model - From Pretraining to PosttrainingLlama-Nemotron:Efficient Reasoning…Llama-Nemotron: Efficient Reasoning ModelsDoes ReinforcementLearning Really…Does Reinforcement Learning Really Incentivize Reasoning Capacity in LLMs Beyond the Base Model?MegaMath: Pushing theLimits of Open Math…MegaMath: Pushing the Limits of Open Math CorporaAceReason-Nemotron:Advancing Math and Code…AceReason-Nemotron: Advancing Math and Code Reasoning through Reinforcement LearningOlmix: A Framework forData Mixing Throughout…Olmix: A Framework for Data Mixing Throughout LM DevelopmentMarco-MoE: OpenMultilingual…Marco-MoE: Open Multilingual Mixture-of-Expert Language Models with Efficient UpcyclingMaximum LikelihoodReinforcement LearningMaximum Likelihood Reinforcement LearningReasonXL: Shifting LLMReasoning Language…ReasonXL: Shifting LLM Reasoning Language Without Sacrificing PerformancemSFT: Addressing DatasetMixtures Overfitting…mSFT: Addressing Dataset Mixtures Overfitting Heterogeneously in Multi-task SFTZAYA1-8B TechnicalReportZAYA1-8B Technical ReportEndless Terminals:Scaling RL Environments…Endless Terminals: Scaling RL Environments for Terminal AgentsMergeMix: OptimizingMid-Training Data…MergeMix: Optimizing Mid-Training Data Mixtures via Learnable Model MergingReasoning Shift: HowContext Silently…Reasoning Shift: How Context Silently Shortens LLM ReasoningSoft Contamination MeansBenchmarks Test Shallow…Soft Contamination Means Benchmarks Test Shallow GeneralizationUnifying Group-Relativeand Self-Distillation…Unifying Group-Relative and Self-Distillation Policy Optimization via Sample RoutingEffective Distillationto Hybrid xLSTM…Effective Distillation to Hybrid xLSTM ArchitecturesOlmo 3Olmo 3Earlier referencesFocus paperCiting papersOlderNewer

Click a node to pin it, click the empty canvas to go back to this paper, or hover to preview. Open a node’s page from its title.