litlas
ログインアカウント作成

gpt-oss-120b & gpt-oss-20b Model Card

著者: OpenAI, :, Agarwal, Sandhini, Ahmad, Lama, Ai, Jason, Altman, Sam, Applebaum, Andy, Arbus, Edwin, Arora, Rahul K., Bai, Yu, Baker, Bowen, Bao, Haiming, Barak, Boaz, Bennett, Ally, Bertao, Tyler, Brett, Nivedita, Brevdo, Eugene, Brockman, Greg, Bubeck, Sebastien, Chang, Che, Chen, Kai, Chen, Mark, Cheung, Enoch, Clark, Aidan, Cook, Dan, Dukhan, Marat, Dvorak, Casey, Fives, Kevin, Fomenko, Vlad, Garipov, Timur ほか 97 名 - arXiv 2025 被引用: 1,242

We present gpt-oss-120b and gpt-oss-20b, two open-weight reasoning models that push the frontier of accuracy and inference cost. The models use an efficient mixture-of-expert transformer architecture and are trained using large-scale distillation and reinforcement learning. We optimize the models to have strong agentic capabilities (deep research browsing, python tool use, and support for developer-provided functions), all while using a rendered chat format that enables clear instruction following and role delineation. Both models achieve strong results on benchmarks ranging from mathematics, coding, and safety. We release the model weights, inference implementations, tool environments, and tokenizers under an Apache 2.0 license to enable broad use and further research.

Generating LongSequences with Sparse…Generating Long Sequences with Sparse TransformersFast TransformerDecoding: One Write-Hea…Fast Transformer Decoding: One Write-Head is All You NeedGLU Variants ImproveTransformerGLU Variants Improve TransformerMeasuring MassiveMultitask Language…Measuring Massive Multitask Language UnderstandingGShard: Scaling GiantModels with Conditional…GShard: Scaling Giant Models with Conditional Computation and Automatic ShardingGPT-4o System CardGPT-4o System Cardτ-bench: A Benchmark forTool-Agent-User…τ-bench: A Benchmark for Tool-Agent-User Interaction in Real-World DomainsEfficient StreamingLanguage Models with…Efficient Streaming Language Models with Attention SinksYaRN: Efficient ContextWindow Extension of…YaRN: Efficient Context Window Extension of Large Language ModelsHumanity's Last ExamHumanity's Last ExamDeliberative Alignment:Reasoning Enables Safer…Deliberative Alignment: Reasoning Enables Safer Language ModelsHealthBench: EvaluatingLarge Language Models…HealthBench: Evaluating Large Language Models Towards Improved Human HealthEvaluating LargeLanguage Models in…Evaluating Large Language Models in Scientific DiscoveryMagentic Marketplace: AnOpen-Source Environment…Magentic Marketplace: An Open-Source Environment for Studying Agentic MarketsMitigating LLMHallucination via…Mitigating LLM Hallucination via Behaviorally Calibrated Reinforcement LearningTrading-R1: FinancialTrading with LLM…Trading-R1: Financial Trading with LLM Reasoning via Reinforcement LearningTritonRL: Training LLMsto Think and Code Trito…TritonRL: Training LLMs to Think and Code Triton Without CheatingQuantifying Risks inMulti-turn Conversation…Quantifying Risks in Multi-turn Conversation with Large Language ModelsBag of Tricks forSubverting…Bag of Tricks for Subverting Reasoning-based Safety GuardrailsInteractScience:Programmatic and…InteractScience: Programmatic and Visually-Grounded Evaluation of Interactive Scientific Demonstration Code GenerationOn Calibration of LargeLanguage Models: From…On Calibration of Large Language Models: From Response To CapabilityProbing the Trajectoriesof Reasoning Traces in…Probing the Trajectories of Reasoning Traces in Large Language ModelsP1-VL: Bridging VisualPerception and…P1-VL: Bridging Visual Perception and Scientific Reasoning in Physics OlympiadsRethinking LanguageModel Scaling under…Rethinking Language Model Scaling under Transferable Hypersphere Optimizationgpt-oss-120b &gpt-oss-20b Model Cardgpt-oss-120b & gpt-oss-20b Model Card過去の参考文献中心の論文この論文を引用する論文古い新しい

ノードをクリックするとフォーカスを固定、空白をクリックすると本論文に戻ります。ホバーで一時的にプレビューできます。各ノードのページはタイトルから開けます。