{"work":{"id":"07c85cc5-4086-4abc-823b-6d0f4ff784d0","openalex_id":null,"doi":null,"arxiv_id":"2512.02556","raw_key":null,"title":"DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models","authors":null,"authors_text":"DeepSeek-AI, Aixin Liu, Aoxue Mei, Bangcai Lin, Bing Xue, Bingxuan Wang","year":2025,"venue":"cs.CL","abstract":"We introduce DeepSeek-V3.2, a model that harmonizes high computational efficiency with superior reasoning and agent performance. The key technical breakthroughs of DeepSeek-V3.2 are as follows: (1) DeepSeek Sparse Attention (DSA): We introduce DSA, an efficient attention mechanism that substantially reduces computational complexity while preserving model performance in long-context scenarios. (2) Scalable Reinforcement Learning Framework: By implementing a robust reinforcement learning protocol and scaling post-training compute, DeepSeek-V3.2 performs comparably to GPT-5. Notably, our high-compute variant, DeepSeek-V3.2-Speciale, surpasses GPT-5 and exhibits reasoning proficiency on par with Gemini-3.0-Pro, achieving gold-medal performance in both the 2025 International Mathematical Olympiad (IMO) and the International Olympiad in Informatics (IOI). (3) Large-Scale Agentic Task Synthesis Pipeline: To integrate reasoning into tool-use scenarios, we developed a novel synthesis pipeline that systematically generates training data at scale. This methodology facilitates scalable agentic post-training, yielding substantial improvements in generalization and instruction-following robustness within complex, interactive environments.","external_url":"https://arxiv.org/abs/2512.02556","cited_by_count":null,"metadata_source":"pith","metadata_fetched_at":"2026-05-14T23:08:14.536209+00:00","pith_arxiv_id":"2512.02556","created_at":"2026-05-09T00:14:27.052088+00:00","updated_at":"2026-05-14T23:08:14.536209+00:00","title_quality_ok":true,"display_title":"DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models","render_title":"DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models"},"hub":{"state":{"work_id":"07c85cc5-4086-4abc-823b-6d0f4ff784d0","tier":"super_hub","tier_reason":"100+ Pith inbound or 10,000+ external citations","pith_inbound_count":126,"external_cited_by_count":null,"distinct_field_count":18,"first_pith_cited_at":"2026-01-06T07:31:47+00:00","last_pith_cited_at":"2026-05-13T08:05:16+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":"2026-05-14T23:06:18.210070+00:00","tier_text":"super_hub"},"tier":"super_hub","role_counts":[{"context_role":"background","n":2}],"polarity_counts":[{"context_polarity":"background","n":2}],"runs":{"ask_index":{"job_type":"ask_index","status":"succeeded","result":{"title":"DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models","claims":[{"claim_text":"We introduce DeepSeek-V3.2, a model that harmonizes high computational efficiency with superior reasoning and agent performance. The key technical breakthroughs of DeepSeek-V3.2 are as follows: (1) DeepSeek Sparse Attention (DSA): We introduce DSA, an efficient attention mechanism that substantially reduces computational complexity while preserving model performance in long-context scenarios. (2) Scalable Reinforcement Learning Framework: By implementing a robust reinforcement learning protocol and scaling post-training compute, DeepSeek-V3.2 performs comparably to GPT-5. Notably, our high-com","claim_type":"abstract","evidence_strength":"source_metadata"}],"why_cited":"Pith tracks DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models because it crossed a citation-hub threshold.","role_counts":[]},"error":null,"updated_at":"2026-05-14T00:54:01.698754+00:00"},"author_expand":{"job_type":"author_expand","status":"succeeded","result":{"authors_linked":[{"id":"b3d3bc38-c7e6-4554-ab1a-a4b32a8299c8","orcid":null,"display_name":"DeepSeek-AI"},{"id":"5ca1d3e9-abe6-434e-91cd-b96f50c305c4","orcid":null,"display_name":"Aixin Liu"},{"id":"64e64001-715d-4aa5-81ec-a7550e627703","orcid":null,"display_name":"Aoxue Mei"},{"id":"e3e6c5af-0098-4799-8c92-809a640684c6","orcid":null,"display_name":"Bangcai Lin"},{"id":"2910ae2c-4464-4656-858c-31b2bca909fb","orcid":null,"display_name":"Bing Xue"},{"id":"1319a98c-3288-417b-86b2-19490b8cdcb2","orcid":null,"display_name":"Bingxuan Wang"}]},"error":null,"updated_at":"2026-05-14T00:54:00.640719+00:00"},"context_extract":{"job_type":"context_extract","status":"succeeded","result":{"enqueued_papers":25},"error":null,"updated_at":"2026-05-14T00:44:10.948280+00:00"},"graph_features":{"job_type":"graph_features","status":"succeeded","result":{"co_cited":[{"title":"Qwen3 Technical Report","work_id":"25a4e30c-1232-48e7-9925-02fa12ba7c9e","shared_citers":64},{"title":"Kimi K2.5: Visual Agentic Intelligence","work_id":"d690be8f-5d53-49b0-b1e7-79668eb8fcdb","shared_citers":24},{"title":"DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models","work_id":"c5006563-f3ec-438a-9e35-b7b484f34828","shared_citers":20},{"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":16},{"title":"GLM-5: from Vibe Coding to Agentic Engineering","work_id":"ad29b1a2-bf77-46b3-9ead-fb62b1d2c6fe","shared_citers":16},{"title":"gpt-oss-120b & gpt-oss-20b Model Card","work_id":"178c1f7e-4f19-4392-a45d-45a6dfa88ead","shared_citers":15},{"title":"DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning","work_id":"e6b75ad5-2877-4168-97c8-710407094d20","shared_citers":14},{"title":"Kimi K2: Open Agentic Intelligence","work_id":"7f18284c-12d3-4137-bea1-1da97e8cf3c1","shared_citers":14},{"title":"OpenAI GPT-5 System Card","work_id":"ca87689a-0d29-4476-b504-b65dbbb08af4","shared_citers":14},{"title":"The Llama 3 Herd of Models","work_id":"1549a635-88af-4ac1-acfe-51ae7bb53345","shared_citers":14},{"title":"DeepSeek-V3 Technical Report","work_id":"57d2791d-2219-4c31-a077-afc04b12a75c","shared_citers":13},{"title":"GPT-4 Technical Report","work_id":"b928e041-6991-4c08-8c81-0359e4097c7b","shared_citers":13},{"title":"Evaluating Large Language Models Trained on Code","work_id":"042493e9-b26f-4b4e-bbde-382072ca9b08","shared_citers":12},{"title":"DAPO: An Open-Source LLM Reinforcement Learning System at Scale","work_id":"64019d00-0b11-4bbd-b173-b46c8fad0157","shared_citers":11},{"title":"Qwen3-VL Technical Report","work_id":"1fe243aa-e3c0-4da6-b391-4cbcfc88d5c0","shared_citers":11},{"title":"Training Verifiers to Solve Math Word Problems","work_id":"acab1aa8-b4d6-40e0-a3ee-25341701dca2","shared_citers":11},{"title":"LiveCodeBench: Holistic and Contamination Free Evaluation of Large Language Models for Code","work_id":"ea9e51ce-1e75-4182-92d8-4d25f70d2ee4","shared_citers":10},{"title":"Proximal Policy Optimization Algorithms","work_id":"240c67fe-d14d-4520-91c1-38a4e272ca19","shared_citers":10},{"title":"GPT-4o System Card","work_id":"f37bf1c7-4964-4e56-9762-d20da8d9009f","shared_citers":8},{"title":"SWE-bench: Can Language Models Resolve Real-World GitHub Issues?","work_id":"d0effe15-a689-441a-8e3f-ea35f1c4e4b1","shared_citers":8},{"title":"Gemini: A Family of Highly Capable Multimodal Models","work_id":"83f7c85b-3f11-450f-ac0c-64d9745220b2","shared_citers":7},{"title":"Longformer: The Long-Document Transformer","work_id":"abea7a44-6668-4de7-aab6-f53a6e5aa088","shared_citers":7},{"title":"Terminal-Bench: Benchmarking Agents on Hard, Realistic Tasks in Command Line Interfaces","work_id":"0624be05-1d97-4fd6-8300-b04b8a3ab04b","shared_citers":7},{"title":"BrowseComp: A Simple Yet Challenging Benchmark for Browsing Agents","work_id":"25adb508-d97c-49d6-ae43-7a70c2478a34","shared_citers":6}],"time_series":[{"n":119,"year":2026}]},"error":null,"updated_at":"2026-05-14T00:54:06.202649+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":"2026-05-14T00:44:15.177114+00:00"},"role_polarity":{"job_type":"role_polarity","status":"succeeded","result":{"title":"DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models","claims":[{"claim_text":"We introduce DeepSeek-V3.2, a model that harmonizes high computational efficiency with superior reasoning and agent performance. The key technical breakthroughs of DeepSeek-V3.2 are as follows: (1) DeepSeek Sparse Attention (DSA): We introduce DSA, an efficient attention mechanism that substantially reduces computational complexity while preserving model performance in long-context scenarios. (2) Scalable Reinforcement Learning Framework: By implementing a robust reinforcement learning protocol and scaling post-training compute, DeepSeek-V3.2 performs comparably to GPT-5. Notably, our high-com","claim_type":"abstract","evidence_strength":"source_metadata"}],"why_cited":"Pith tracks DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models because it crossed a citation-hub threshold.","role_counts":[]},"error":null,"updated_at":"2026-05-14T00:54:01.705942+00:00"},"summary_claims":{"job_type":"summary_claims","status":"succeeded","result":{"title":"DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models","claims":[{"claim_text":"We introduce DeepSeek-V3.2, a model that harmonizes high computational efficiency with superior reasoning and agent performance. The key technical breakthroughs of DeepSeek-V3.2 are as follows: (1) DeepSeek Sparse Attention (DSA): We introduce DSA, an efficient attention mechanism that substantially reduces computational complexity while preserving model performance in long-context scenarios. (2) Scalable Reinforcement Learning Framework: By implementing a robust reinforcement learning protocol and scaling post-training compute, DeepSeek-V3.2 performs comparably to GPT-5. Notably, our high-com","claim_type":"abstract","evidence_strength":"source_metadata"}],"why_cited":"Pith tracks DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models because it crossed a citation-hub threshold.","role_counts":[]},"error":null,"updated_at":"2026-05-14T00:54:01.695460+00:00"}},"summary":{"title":"DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models","claims":[{"claim_text":"We introduce DeepSeek-V3.2, a model that harmonizes high computational efficiency with superior reasoning and agent performance. The key technical breakthroughs of DeepSeek-V3.2 are as follows: (1) DeepSeek Sparse Attention (DSA): We introduce DSA, an efficient attention mechanism that substantially reduces computational complexity while preserving model performance in long-context scenarios. (2) Scalable Reinforcement Learning Framework: By implementing a robust reinforcement learning protocol and scaling post-training compute, DeepSeek-V3.2 performs comparably to GPT-5. Notably, our high-com","claim_type":"abstract","evidence_strength":"source_metadata"}],"why_cited":"Pith tracks DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models because it crossed a citation-hub threshold.","role_counts":[]},"graph":{"co_cited":[{"title":"Qwen3 Technical Report","work_id":"25a4e30c-1232-48e7-9925-02fa12ba7c9e","shared_citers":64},{"title":"Kimi K2.5: Visual Agentic Intelligence","work_id":"d690be8f-5d53-49b0-b1e7-79668eb8fcdb","shared_citers":24},{"title":"DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models","work_id":"c5006563-f3ec-438a-9e35-b7b484f34828","shared_citers":20},{"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":16},{"title":"GLM-5: from Vibe Coding to Agentic Engineering","work_id":"ad29b1a2-bf77-46b3-9ead-fb62b1d2c6fe","shared_citers":16},{"title":"gpt-oss-120b & gpt-oss-20b Model Card","work_id":"178c1f7e-4f19-4392-a45d-45a6dfa88ead","shared_citers":15},{"title":"DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning","work_id":"e6b75ad5-2877-4168-97c8-710407094d20","shared_citers":14},{"title":"Kimi K2: Open Agentic Intelligence","work_id":"7f18284c-12d3-4137-bea1-1da97e8cf3c1","shared_citers":14},{"title":"OpenAI GPT-5 System Card","work_id":"ca87689a-0d29-4476-b504-b65dbbb08af4","shared_citers":14},{"title":"The Llama 3 Herd of Models","work_id":"1549a635-88af-4ac1-acfe-51ae7bb53345","shared_citers":14},{"title":"DeepSeek-V3 Technical Report","work_id":"57d2791d-2219-4c31-a077-afc04b12a75c","shared_citers":13},{"title":"GPT-4 Technical Report","work_id":"b928e041-6991-4c08-8c81-0359e4097c7b","shared_citers":13},{"title":"Evaluating Large Language Models Trained on Code","work_id":"042493e9-b26f-4b4e-bbde-382072ca9b08","shared_citers":12},{"title":"DAPO: An Open-Source LLM Reinforcement Learning System at Scale","work_id":"64019d00-0b11-4bbd-b173-b46c8fad0157","shared_citers":11},{"title":"Qwen3-VL Technical Report","work_id":"1fe243aa-e3c0-4da6-b391-4cbcfc88d5c0","shared_citers":11},{"title":"Training Verifiers to Solve Math Word Problems","work_id":"acab1aa8-b4d6-40e0-a3ee-25341701dca2","shared_citers":11},{"title":"LiveCodeBench: Holistic and Contamination Free Evaluation of Large Language Models for Code","work_id":"ea9e51ce-1e75-4182-92d8-4d25f70d2ee4","shared_citers":10},{"title":"Proximal Policy Optimization Algorithms","work_id":"240c67fe-d14d-4520-91c1-38a4e272ca19","shared_citers":10},{"title":"GPT-4o System Card","work_id":"f37bf1c7-4964-4e56-9762-d20da8d9009f","shared_citers":8},{"title":"SWE-bench: Can Language Models Resolve Real-World GitHub Issues?","work_id":"d0effe15-a689-441a-8e3f-ea35f1c4e4b1","shared_citers":8},{"title":"Gemini: A Family of Highly Capable Multimodal Models","work_id":"83f7c85b-3f11-450f-ac0c-64d9745220b2","shared_citers":7},{"title":"Longformer: The Long-Document Transformer","work_id":"abea7a44-6668-4de7-aab6-f53a6e5aa088","shared_citers":7},{"title":"Terminal-Bench: Benchmarking Agents on Hard, Realistic Tasks in Command Line Interfaces","work_id":"0624be05-1d97-4fd6-8300-b04b8a3ab04b","shared_citers":7},{"title":"BrowseComp: A Simple Yet Challenging Benchmark for Browsing Agents","work_id":"25adb508-d97c-49d6-ae43-7a70c2478a34","shared_citers":6}],"time_series":[{"n":119,"year":2026}]},"authors":[{"id":"5ca1d3e9-abe6-434e-91cd-b96f50c305c4","orcid":null,"display_name":"Aixin Liu","source":"manual","import_confidence":0.72},{"id":"64e64001-715d-4aa5-81ec-a7550e627703","orcid":null,"display_name":"Aoxue Mei","source":"manual","import_confidence":0.72},{"id":"e3e6c5af-0098-4799-8c92-809a640684c6","orcid":null,"display_name":"Bangcai Lin","source":"manual","import_confidence":0.72},{"id":"1319a98c-3288-417b-86b2-19490b8cdcb2","orcid":null,"display_name":"Bingxuan Wang","source":"manual","import_confidence":0.72},{"id":"2910ae2c-4464-4656-858c-31b2bca909fb","orcid":null,"display_name":"Bing Xue","source":"manual","import_confidence":0.72},{"id":"b3d3bc38-c7e6-4554-ab1a-a4b32a8299c8","orcid":null,"display_name":"DeepSeek-AI","source":"manual","import_confidence":0.72}]}}