{"work":{"id":"68d3c334-0fc9-49e3-b7b0-a69afae933e2","openalex_id":"https://openalex.org/W4405766390","doi":"10.48550/arxiv.2412.16720","arxiv_id":"2412.16720","raw_key":null,"title":"OpenAI o1 System Card","authors":null,"authors_text":"OpenAI, :, Aaron Jaech, Adam Tauman Kalai, Adam Lerer, Adam J. Richardson, Ahmed El-Kishky, A.M. Low, Alec Helyar, Aleksander Mądry","year":2024,"venue":"cs.AI","abstract":"The o1 model series is trained with large-scale reinforcement learning to reason using chain of thought. These advanced reasoning capabilities provide new avenues for improving the safety and robustness of our models. In particular, our models can reason about our safety policies in context when responding to potentially unsafe prompts, through deliberative alignment. This leads to state-of-the-art performance on certain benchmarks for risks such as generating illicit advice, choosing stereotyped responses, and succumbing to known jailbreaks. Training models to incorporate a chain of thought before answering has the potential to unlock substantial benefits, while also increasing potential risks that stem from heightened intelligence. Our results underscore the need for building robust alignment methods, extensively stress-testing their efficacy, and maintaining meticulous risk management protocols. This report outlines the safety work carried out for the OpenAI o1 and OpenAI o1-mini models, including safety evaluations, external red teaming, and Preparedness Framework evaluations.","external_url":"https://arxiv.org/abs/2412.16720","cited_by_count":37,"metadata_source":"pith","metadata_fetched_at":"2026-08-05T02:28:24.338817+00:00","pith_arxiv_id":"2412.16720","created_at":"2026-05-09T03:25:03.252771+00:00","updated_at":"2026-08-05T02:49:54.815029+00:00","title_quality_ok":false,"display_title":"OpenAI o1 System Card","render_title":"OpenAI o1 System Card"},"hub":{"state":{"work_id":"68d3c334-0fc9-49e3-b7b0-a69afae933e2","tier":"super_hub","tier_reason":"100+ Pith inbound or 10,000+ external citations","pith_inbound_count":493,"external_cited_by_count":37,"distinct_field_count":21,"first_pith_cited_at":"2024-10-05T06:04:19+00:00","last_pith_cited_at":"2026-07-09T17:30:49+00:00","author_build_status":"needed","summary_status":"needed","contexts_status":"needed","graph_status":"needed","ask_index_status":"needed","reader_status":"not_needed","recognition_status":"not_needed","updated_at":"2026-08-22T16:49:30.723283+00:00","tier_text":"super_hub"},"tier":"super_hub","role_counts":[{"context_role":"background","n":100},{"context_role":"baseline","n":5},{"context_role":"method","n":5},{"context_role":"dataset","n":2},{"context_role":"other","n":1}],"polarity_counts":[{"context_polarity":"background","n":96},{"context_polarity":"baseline","n":5},{"context_polarity":"unclear","n":5},{"context_polarity":"use_method","n":4},{"context_polarity":"use_dataset","n":2},{"context_polarity":"support","n":1}],"runs":{"ask_index":{"job_type":"ask_index","status":"succeeded","result":{"title":"OpenAI o1 System Card","claims":[{"claim_text":"The o1 model series is trained with large-scale reinforcement learning to reason using chain of thought. These advanced reasoning capabilities provide new avenues for improving the safety and robustness of our models. In particular, our models can reason about our safety policies in context when responding to potentially unsafe prompts, through deliberative alignment. This leads to state-of-the-art performance on certain benchmarks for risks such as generating illicit advice, choosing stereotyped responses, and succumbing to known jailbreaks. Training models to incorporate a chain of thought b","claim_type":"abstract","evidence_strength":"source_metadata"}],"why_cited":"Pith tracks OpenAI o1 System Card because it crossed a citation-hub threshold.","role_counts":[]},"error":null,"updated_at":"2026-05-13T21:13:34.435005+00:00"},"author_expand":{"job_type":"author_expand","status":"succeeded","result":{"authors_linked":[]},"error":null,"updated_at":"2026-05-13T21:13:36.962597+00:00"},"context_extract":{"job_type":"context_extract","status":"succeeded","result":{"enqueued_papers":25},"error":null,"updated_at":"2026-05-13T21:03:42.140243+00:00"},"graph_features":{"job_type":"graph_features","status":"succeeded","result":{"co_cited":[{"title":"DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning","work_id":"e6b75ad5-2877-4168-97c8-710407094d20","shared_citers":107},{"title":"DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models","work_id":"c5006563-f3ec-438a-9e35-b7b484f34828","shared_citers":92},{"title":"Qwen3 Technical Report","work_id":"25a4e30c-1232-48e7-9925-02fa12ba7c9e","shared_citers":67},{"title":"Proximal Policy Optimization Algorithms","work_id":"240c67fe-d14d-4520-91c1-38a4e272ca19","shared_citers":55},{"title":"DAPO: An Open-Source LLM Reinforcement Learning System at Scale","work_id":"64019d00-0b11-4bbd-b173-b46c8fad0157","shared_citers":47},{"title":"GPT-4 Technical Report","work_id":"b928e041-6991-4c08-8c81-0359e4097c7b","shared_citers":35},{"title":"The Llama 3 Herd of Models","work_id":"1549a635-88af-4ac1-acfe-51ae7bb53345","shared_citers":32},{"title":"GPT-4o System Card","work_id":"f37bf1c7-4964-4e56-9762-d20da8d9009f","shared_citers":31},{"title":"Training Verifiers to Solve Math Word Problems","work_id":"acab1aa8-b4d6-40e0-a3ee-25341701dca2","shared_citers":31},{"title":"Qwen2.5-VL Technical Report","work_id":"69dffacb-bfe8-442d-be86-48624c60426f","shared_citers":30},{"title":"Kimi k1.5: Scaling Reinforcement Learning with LLMs","work_id":"bff96ab1-bd6a-4585-be23-74fdb51969c7","shared_citers":28},{"title":"Qwen2.5 Technical Report","work_id":"d8432992-4980-4a81-85c7-9fa2c2b87f85","shared_citers":28},{"title":"Measuring Mathematical Problem Solving With the MATH Dataset","work_id":"50652ac6-fb7c-4675-a2c2-159c241feb17","shared_citers":25},{"title":"Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities","work_id":"008df105-2fdd-45d8-857a-8e35868aecb6","shared_citers":24},{"title":"Group Sequence Policy Optimization","work_id":"3a98b53b-9f52-4d95-adf7-89353c0a9a65","shared_citers":22},{"title":"Understanding R1-Zero-Like Training: A Critical Perspective","work_id":"ec354f3b-9484-4a0c-94c8-92d4d0260835","shared_citers":21},{"title":"DeepSeek-V3 Technical Report","work_id":"57d2791d-2219-4c31-a077-afc04b12a75c","shared_citers":19},{"title":"Evaluating Large Language Models Trained on Code","work_id":"042493e9-b26f-4b4e-bbde-382072ca9b08","shared_citers":19},{"title":"Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters","work_id":"a8d50b24-bdf5-46ed-bc4f-2927dfd81f1d","shared_citers":18},{"title":"HybridFlow: A Flexible and Efficient RLHF Framework","work_id":"7eb9c9f4-b322-4bba-8011-09ff8d6ad801","shared_citers":15},{"title":"Tulu 3: Pushing Frontiers in Open Language Model Post-Training","work_id":"28c9dbea-056a-48c2-8000-85f809827e45","shared_citers":15},{"title":"Qwen2.5-Math Technical Report: Toward Mathematical Expert Model via Self-Improvement","work_id":"a097c5d4-6d32-46ee-9826-57d532bbfc9c","shared_citers":14},{"title":"OpenAI GPT-5 System Card","work_id":"ca87689a-0d29-4476-b504-b65dbbb08af4","shared_citers":13},{"title":"Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution","work_id":"8abcfe4f-e0fb-44b7-9123-448fac95f90a","shared_citers":13}],"time_series":[{"n":17,"year":2025},{"n":153,"year":2026}]},"error":null,"updated_at":"2026-05-13T21:13:34.511489+00:00"},"identity_refresh":{"job_type":"identity_refresh","status":"succeeded","result":{"fixed":1,"items":[{"title":"Qwen3 Technical Report","work_id":"25a4e30c-1232-48e7-9925-02fa12ba7c9e","resolver":"local_arxiv","confidence":0.98,"old_work_id":"25a4e30c-1232-48e7-9925-02fa12ba7c9e"}],"errors":[],"attempted":1},"error":null,"updated_at":"2026-05-13T21:03:39.912711+00:00"},"role_polarity":{"job_type":"role_polarity","status":"succeeded","result":{"title":"OpenAI o1 System Card","claims":[{"claim_text":"The o1 model series is trained with large-scale reinforcement learning to reason using chain of thought. These advanced reasoning capabilities provide new avenues for improving the safety and robustness of our models. In particular, our models can reason about our safety policies in context when responding to potentially unsafe prompts, through deliberative alignment. This leads to state-of-the-art performance on certain benchmarks for risks such as generating illicit advice, choosing stereotyped responses, and succumbing to known jailbreaks. Training models to incorporate a chain of thought b","claim_type":"abstract","evidence_strength":"source_metadata"}],"why_cited":"Pith tracks OpenAI o1 System Card because it crossed a citation-hub threshold.","role_counts":[]},"error":null,"updated_at":"2026-05-13T21:03:37.289668+00:00"},"summary_claims":{"job_type":"summary_claims","status":"succeeded","result":{"title":"OpenAI o1 System Card","claims":[{"claim_text":"The o1 model series is trained with large-scale reinforcement learning to reason using chain of thought. These advanced reasoning capabilities provide new avenues for improving the safety and robustness of our models. In particular, our models can reason about our safety policies in context when responding to potentially unsafe prompts, through deliberative alignment. This leads to state-of-the-art performance on certain benchmarks for risks such as generating illicit advice, choosing stereotyped responses, and succumbing to known jailbreaks. Training models to incorporate a chain of thought b","claim_type":"abstract","evidence_strength":"source_metadata"}],"why_cited":"Pith tracks OpenAI o1 System Card because it crossed a citation-hub threshold.","role_counts":[]},"error":null,"updated_at":"2026-05-13T21:03:41.080462+00:00"}},"summary":{"title":"OpenAI o1 System Card","claims":[{"claim_text":"The o1 model series is trained with large-scale reinforcement learning to reason using chain of thought. These advanced reasoning capabilities provide new avenues for improving the safety and robustness of our models. In particular, our models can reason about our safety policies in context when responding to potentially unsafe prompts, through deliberative alignment. This leads to state-of-the-art performance on certain benchmarks for risks such as generating illicit advice, choosing stereotyped responses, and succumbing to known jailbreaks. Training models to incorporate a chain of thought b","claim_type":"abstract","evidence_strength":"source_metadata"}],"why_cited":"Pith tracks OpenAI o1 System Card because it crossed a citation-hub threshold.","role_counts":[]},"graph":{"co_cited":[{"title":"DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning","work_id":"e6b75ad5-2877-4168-97c8-710407094d20","shared_citers":107},{"title":"DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models","work_id":"c5006563-f3ec-438a-9e35-b7b484f34828","shared_citers":92},{"title":"Qwen3 Technical Report","work_id":"25a4e30c-1232-48e7-9925-02fa12ba7c9e","shared_citers":67},{"title":"Proximal Policy Optimization Algorithms","work_id":"240c67fe-d14d-4520-91c1-38a4e272ca19","shared_citers":55},{"title":"DAPO: An Open-Source LLM Reinforcement Learning System at Scale","work_id":"64019d00-0b11-4bbd-b173-b46c8fad0157","shared_citers":47},{"title":"GPT-4 Technical Report","work_id":"b928e041-6991-4c08-8c81-0359e4097c7b","shared_citers":35},{"title":"The Llama 3 Herd of Models","work_id":"1549a635-88af-4ac1-acfe-51ae7bb53345","shared_citers":32},{"title":"GPT-4o System Card","work_id":"f37bf1c7-4964-4e56-9762-d20da8d9009f","shared_citers":31},{"title":"Training Verifiers to Solve Math Word Problems","work_id":"acab1aa8-b4d6-40e0-a3ee-25341701dca2","shared_citers":31},{"title":"Qwen2.5-VL Technical Report","work_id":"69dffacb-bfe8-442d-be86-48624c60426f","shared_citers":30},{"title":"Kimi k1.5: Scaling Reinforcement Learning with LLMs","work_id":"bff96ab1-bd6a-4585-be23-74fdb51969c7","shared_citers":28},{"title":"Qwen2.5 Technical Report","work_id":"d8432992-4980-4a81-85c7-9fa2c2b87f85","shared_citers":28},{"title":"Measuring Mathematical Problem Solving With the MATH Dataset","work_id":"50652ac6-fb7c-4675-a2c2-159c241feb17","shared_citers":25},{"title":"Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities","work_id":"008df105-2fdd-45d8-857a-8e35868aecb6","shared_citers":24},{"title":"Group Sequence Policy Optimization","work_id":"3a98b53b-9f52-4d95-adf7-89353c0a9a65","shared_citers":22},{"title":"Understanding R1-Zero-Like Training: A Critical Perspective","work_id":"ec354f3b-9484-4a0c-94c8-92d4d0260835","shared_citers":21},{"title":"DeepSeek-V3 Technical Report","work_id":"57d2791d-2219-4c31-a077-afc04b12a75c","shared_citers":19},{"title":"Evaluating Large Language Models Trained on Code","work_id":"042493e9-b26f-4b4e-bbde-382072ca9b08","shared_citers":19},{"title":"Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters","work_id":"a8d50b24-bdf5-46ed-bc4f-2927dfd81f1d","shared_citers":18},{"title":"HybridFlow: A Flexible and Efficient RLHF Framework","work_id":"7eb9c9f4-b322-4bba-8011-09ff8d6ad801","shared_citers":15},{"title":"Tulu 3: Pushing Frontiers in Open Language Model Post-Training","work_id":"28c9dbea-056a-48c2-8000-85f809827e45","shared_citers":15},{"title":"Qwen2.5-Math Technical Report: Toward Mathematical Expert Model via Self-Improvement","work_id":"a097c5d4-6d32-46ee-9826-57d532bbfc9c","shared_citers":14},{"title":"OpenAI GPT-5 System Card","work_id":"ca87689a-0d29-4476-b504-b65dbbb08af4","shared_citers":13},{"title":"Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution","work_id":"8abcfe4f-e0fb-44b7-9123-448fac95f90a","shared_citers":13}],"time_series":[{"n":17,"year":2025},{"n":153,"year":2026}]},"authors":[]}}