{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2025:DD2JVSLQM3CM3D5QV5MIRRQTNY","short_pith_number":"pith:DD2JVSLQ","canonical_record":{"source":{"id":"2508.10925","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-08-08T19:24:38Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"bb7b42f5055e7b4fc5e5e22b75da7aae00025a373a573422058f2fd7342fce08","abstract_canon_sha256":"e163e6febcae44ec88546116a93ce4b987d40bc98f4d9887a12b9907100fbef1"},"schema_version":"1.0"},"canonical_sha256":"18f49ac97066c4cd8fb0af5888c6136e058f6d471017344d133edadf03aa70d1","source":{"kind":"arxiv","id":"2508.10925","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2508.10925","created_at":"2026-07-05T11:54:15Z"},{"alias_kind":"arxiv_version","alias_value":"2508.10925v1","created_at":"2026-07-05T11:54:15Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2508.10925","created_at":"2026-07-05T11:54:15Z"},{"alias_kind":"pith_short_12","alias_value":"DD2JVSLQM3CM","created_at":"2026-07-05T11:54:15Z"},{"alias_kind":"pith_short_16","alias_value":"DD2JVSLQM3CM3D5Q","created_at":"2026-07-05T11:54:15Z"},{"alias_kind":"pith_short_8","alias_value":"DD2JVSLQ","created_at":"2026-07-05T11:54:15Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2025:DD2JVSLQM3CM3D5QV5MIRRQTNY","target":"record","payload":{"canonical_record":{"source":{"id":"2508.10925","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-08-08T19:24:38Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"bb7b42f5055e7b4fc5e5e22b75da7aae00025a373a573422058f2fd7342fce08","abstract_canon_sha256":"e163e6febcae44ec88546116a93ce4b987d40bc98f4d9887a12b9907100fbef1"},"schema_version":"1.0"},"canonical_sha256":"18f49ac97066c4cd8fb0af5888c6136e058f6d471017344d133edadf03aa70d1","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:54:15.454537Z","signature_b64":"78t509B3BpNlLw4q/AIcDX6V8Mym8L9NI4D63nhRJhrAb9KiduPppjV3qzZjoM5e3edN3GA2ObM6Gj2d+p/aCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"18f49ac97066c4cd8fb0af5888c6136e058f6d471017344d133edadf03aa70d1","last_reissued_at":"2026-07-05T11:54:15.454049Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:54:15.454049Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2508.10925","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T11:54:15Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"a0D/B69Kj8nU8fNX9XMgfyj18EYWPjJF7gwq/TfWfctjdDxWMCHNI49+38F1URXgo4Pln0yYLvuUoVk4UHAACg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-04T09:17:20.339148Z"},"content_sha256":"d0aa9311ad7646a406b8e819dc6dd85441a1b957b366ea7a0de54b42d285ccb9","schema_version":"1.0","event_id":"sha256:d0aa9311ad7646a406b8e819dc6dd85441a1b957b366ea7a0de54b42d285ccb9"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2025:DD2JVSLQM3CM3D5QV5MIRRQTNY","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"gpt-oss-120b & gpt-oss-20b Model Card","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Two open-weight models using mixture-of-expert architecture deliver strong results on math, coding, and safety benchmarks while supporting agentic tool use.","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Adam Goucher, Aidan Clark, Aidan McLaughlin, Alec Helyar, Alexander Neitz, Alex Nichol, Alex Paino, Ally Bennett, Amy Wendling, Andy Applebaum, Ashley Pantuliano, Boaz Barak, Bob Rotsted, Bowen Baker, Brian Zhang, Carolina Paz, Casey Dvorak, Cedric Whitney, Che Chang, Chris Koch, Chris Lu, Cyril Zhang, Dana Palmie, Dan Cook, Dane Stuckey, David Robinson, Dmitry Pimenov, Dominik Kundel, D. Sculley, Eddie Zhang, Edwin Arbus, Elaine Ya Le, Elizabeth Proehl, Enoch Cheung, Eric Wallace, Eugene Brevdo, Filippo Raso, Foivos Tsimpourlas, Gaby Raila, Giambattista Parascandolo, Gideon Myles, Greg Brockman, Guillaume Leclerc, Hadi Salman, Haiming Bao, Haitang Hu, Hannah Wong, Harshit Sikchi, Hongyu Ren, Huida Qiu, Irina Kofman, Jackie Hehir, Jacob Huh, Jakub Pachocki, James Park Lennon, Jason Ai, Jason Kwon, Jiancheng Liu, Ji Lin, Johannes Heidecke, John Hallman, Jongsoo Park, Jordan Liss, Josh McGrath, Kai Chen, Karan Singhal, Katia Gil Guzman, Kendal Simon, Kevin Fives, Kevin Lu, Kevin Weil, Kevin Whinnery, Kimmy Richardson, Kristen Ying, Kristian Georgiev, Lama Ahmad, Leher Pathak, Lily (Xiaoxuan) Liu, Lindsay McCallum, Lin Yang, Ludovic Peran, Lukas Gross, Marat Dukhan, Mario Lezcano-Casado, Mark Chen, Max Schwarzer, Mia Glaese, Michelle Pokrass, Michihiro Yasunaga, Miles Wang, Nikhil Vyas, Nivedita Brett, Olivia Watkins, OpenAI: Sandhini Agarwal, Philippe Tillet, Rahul K. Arora, Romain Huet, Saachi Jain, Sam Altman, Sam Toizer, Scott Lessans, Scott McKinney, Sebastien Bubeck, Shengjia Zhao, Song Mei, Steve Mostovoy, Suvansh Sanjeev, Tarun Gogineni, Timur Garipov, Tong Mu, Tyler Bertao, Vlad Fomenko, Volodymyr Kyrylov, Wenting Zhan, Wojciech Zaremba, Xin Wang, Yang Song, Yuanzhi Li, Yu Bai, Yu Yang, Zach Johnson, Zhiqing Sun, Zhuohan Li, Zoran Martinovic","submitted_at":"2025-08-08T19:24:38Z","abstract_excerpt":"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, "},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"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.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The assumption that the internal training process and benchmark evaluations accurately reflect real-world agentic performance and safety without detailed public evidence or specific scores provided in the abstract.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"OpenAI releases two open-weight reasoning models, gpt-oss-120b and gpt-oss-20b, trained via distillation and RL with claimed strong results on math, coding, and safety benchmarks.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Two open-weight models using mixture-of-expert architecture deliver strong results on math, coding, and safety benchmarks while supporting agentic tool use.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"a0de5bd0001264464f1d70d3d2584602e80bf30942c422985d4beab84542b1fb"},"source":{"id":"2508.10925","kind":"arxiv","version":1},"verdict":{"id":"0baf52b7-c120-45cb-844b-3e9013774d62","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-10T12:18:31.875705Z","strongest_claim":"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.","one_line_summary":"OpenAI releases two open-weight reasoning models, gpt-oss-120b and gpt-oss-20b, trained via distillation and RL with claimed strong results on math, coding, and safety benchmarks.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The assumption that the internal training process and benchmark evaluations accurately reflect real-world agentic performance and safety without detailed public evidence or specific scores provided in the abstract.","pith_extraction_headline":"Two open-weight models using mixture-of-expert architecture deliver strong results on math, coding, and safety benchmarks while supporting agentic tool use."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2508.10925/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":35,"sample":[{"doi":"","year":2017,"title":"Attention is all you need","work_id":"dc501018-e4ae-468a-9b09-0a5942c79ebc","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2017,"title":"Outrageously large neural networks: The sparsely-gated mixture-of-experts layer","work_id":"6ec9184d-c50e-4769-9759-8fc0b9aacdc0","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2006,"title":"GShard: Scaling Giant Models with Conditional Computation and Automatic Sharding","work_id":"52b3c9a6-2a27-45a7-ba2b-ebe4b5bb5a5f","ref_index":3,"cited_arxiv_id":"2006.16668","is_internal_anchor":true},{"doi":"","year":2022,"title":"Glam: Efficient scaling of language models with mixture-of-experts","work_id":"172ee0a2-6c9f-4754-9fe8-ff7de8f51bd1","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"OCP Microscaling Formats (MX) Specification Version 1.0","work_id":"1f9f1ea8-3835-461f-9f2b-73159585fc84","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":35,"snapshot_sha256":"0de0a05ccb0a0207fb3fa85d60c5a748991e5e1519d99a250f94e33589763b72","internal_anchors":14},"formal_canon":{"evidence_count":2,"snapshot_sha256":"c841fd9163a3803ee94c2fbcf8b0684b2f52ffd3a2588df27cacd966a20d7074"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":"0baf52b7-c120-45cb-844b-3e9013774d62"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T11:54:15Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"xIom7ioWlMoLkBUI4n/qVwKt/ZA6PIB+oigqP39MO8jK3S0WXAa+ztuD16vit4T5Vhyeu9/MW/CVQtR27vkrBQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-04T09:17:20.339960Z"},"content_sha256":"877d8b2f9919dc6d38f5f2864edd3c6a590a10c196683ec789043c4e1a9fe849","schema_version":"1.0","event_id":"sha256:877d8b2f9919dc6d38f5f2864edd3c6a590a10c196683ec789043c4e1a9fe849"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/DD2JVSLQM3CM3D5QV5MIRRQTNY/bundle.json","state_url":"https://pith.science/pith/DD2JVSLQM3CM3D5QV5MIRRQTNY/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/DD2JVSLQM3CM3D5QV5MIRRQTNY/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-08-04T09:17:20Z","links":{"resolver":"https://pith.science/pith/DD2JVSLQM3CM3D5QV5MIRRQTNY","bundle":"https://pith.science/pith/DD2JVSLQM3CM3D5QV5MIRRQTNY/bundle.json","state":"https://pith.science/pith/DD2JVSLQM3CM3D5QV5MIRRQTNY/state.json","well_known_bundle":"https://pith.science/.well-known/pith/DD2JVSLQM3CM3D5QV5MIRRQTNY/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:DD2JVSLQM3CM3D5QV5MIRRQTNY","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"e163e6febcae44ec88546116a93ce4b987d40bc98f4d9887a12b9907100fbef1","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-08-08T19:24:38Z","title_canon_sha256":"bb7b42f5055e7b4fc5e5e22b75da7aae00025a373a573422058f2fd7342fce08"},"schema_version":"1.0","source":{"id":"2508.10925","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2508.10925","created_at":"2026-07-05T11:54:15Z"},{"alias_kind":"arxiv_version","alias_value":"2508.10925v1","created_at":"2026-07-05T11:54:15Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2508.10925","created_at":"2026-07-05T11:54:15Z"},{"alias_kind":"pith_short_12","alias_value":"DD2JVSLQM3CM","created_at":"2026-07-05T11:54:15Z"},{"alias_kind":"pith_short_16","alias_value":"DD2JVSLQM3CM3D5Q","created_at":"2026-07-05T11:54:15Z"},{"alias_kind":"pith_short_8","alias_value":"DD2JVSLQ","created_at":"2026-07-05T11:54:15Z"}],"graph_snapshots":[{"event_id":"sha256:877d8b2f9919dc6d38f5f2864edd3c6a590a10c196683ec789043c4e1a9fe849","target":"graph","created_at":"2026-07-05T11:54:15Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"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."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"The assumption that the internal training process and benchmark evaluations accurately reflect real-world agentic performance and safety without detailed public evidence or specific scores provided in the abstract."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"OpenAI releases two open-weight reasoning models, gpt-oss-120b and gpt-oss-20b, trained via distillation and RL with claimed strong results on math, coding, and safety benchmarks."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Two open-weight models using mixture-of-expert architecture deliver strong results on math, coding, and safety benchmarks while supporting agentic tool use."}],"snapshot_sha256":"a0de5bd0001264464f1d70d3d2584602e80bf30942c422985d4beab84542b1fb"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"c841fd9163a3803ee94c2fbcf8b0684b2f52ffd3a2588df27cacd966a20d7074"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2508.10925/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"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, ","authors_text":"Adam Goucher, Aidan Clark, Aidan McLaughlin, Alec Helyar, Alexander Neitz, Alex Nichol, Alex Paino, Ally Bennett, Amy Wendling, Andy Applebaum, Ashley Pantuliano, Boaz Barak, Bob Rotsted, Bowen Baker, Brian Zhang, Carolina Paz, Casey Dvorak, Cedric Whitney, Che Chang, Chris Koch, Chris Lu, Cyril Zhang, Dana Palmie, Dan Cook, Dane Stuckey, David Robinson, Dmitry Pimenov, Dominik Kundel, D. Sculley, Eddie Zhang, Edwin Arbus, Elaine Ya Le, Elizabeth Proehl, Enoch Cheung, Eric Wallace, Eugene Brevdo, Filippo Raso, Foivos Tsimpourlas, Gaby Raila, Giambattista Parascandolo, Gideon Myles, Greg Brockman, Guillaume Leclerc, Hadi Salman, Haiming Bao, Haitang Hu, Hannah Wong, Harshit Sikchi, Hongyu Ren, Huida Qiu, Irina Kofman, Jackie Hehir, Jacob Huh, Jakub Pachocki, James Park Lennon, Jason Ai, Jason Kwon, Jiancheng Liu, Ji Lin, Johannes Heidecke, John Hallman, Jongsoo Park, Jordan Liss, Josh McGrath, Kai Chen, Karan Singhal, Katia Gil Guzman, Kendal Simon, Kevin Fives, Kevin Lu, Kevin Weil, Kevin Whinnery, Kimmy Richardson, Kristen Ying, Kristian Georgiev, Lama Ahmad, Leher Pathak, Lily (Xiaoxuan) Liu, Lindsay McCallum, Lin Yang, Ludovic Peran, Lukas Gross, Marat Dukhan, Mario Lezcano-Casado, Mark Chen, Max Schwarzer, Mia Glaese, Michelle Pokrass, Michihiro Yasunaga, Miles Wang, Nikhil Vyas, Nivedita Brett, Olivia Watkins, OpenAI: Sandhini Agarwal, Philippe Tillet, Rahul K. Arora, Romain Huet, Saachi Jain, Sam Altman, Sam Toizer, Scott Lessans, Scott McKinney, Sebastien Bubeck, Shengjia Zhao, Song Mei, Steve Mostovoy, Suvansh Sanjeev, Tarun Gogineni, Timur Garipov, Tong Mu, Tyler Bertao, Vlad Fomenko, Volodymyr Kyrylov, Wenting Zhan, Wojciech Zaremba, Xin Wang, Yang Song, Yuanzhi Li, Yu Bai, Yu Yang, Zach Johnson, Zhiqing Sun, Zhuohan Li, Zoran Martinovic","cross_cats":["cs.AI"],"headline":"Two open-weight models using mixture-of-expert architecture deliver strong results on math, coding, and safety benchmarks while supporting agentic tool use.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-08-08T19:24:38Z","title":"gpt-oss-120b & gpt-oss-20b Model Card"},"references":{"count":35,"internal_anchors":14,"resolved_work":35,"sample":[{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":1,"title":"Attention is all you need","work_id":"dc501018-e4ae-468a-9b09-0a5942c79ebc","year":2017},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":2,"title":"Outrageously large neural networks: The sparsely-gated mixture-of-experts layer","work_id":"6ec9184d-c50e-4769-9759-8fc0b9aacdc0","year":2017},{"cited_arxiv_id":"2006.16668","doi":"","is_internal_anchor":true,"ref_index":3,"title":"GShard: Scaling Giant Models with Conditional Computation and Automatic Sharding","work_id":"52b3c9a6-2a27-45a7-ba2b-ebe4b5bb5a5f","year":2006},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":4,"title":"Glam: Efficient scaling of language models with mixture-of-experts","work_id":"172ee0a2-6c9f-4754-9fe8-ff7de8f51bd1","year":2022},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":5,"title":"OCP Microscaling Formats (MX) Specification Version 1.0","work_id":"1f9f1ea8-3835-461f-9f2b-73159585fc84","year":2023}],"snapshot_sha256":"0de0a05ccb0a0207fb3fa85d60c5a748991e5e1519d99a250f94e33589763b72"},"source":{"id":"2508.10925","kind":"arxiv","version":1},"verdict":{"created_at":"2026-05-10T12:18:31.875705Z","id":"0baf52b7-c120-45cb-844b-3e9013774d62","model_set":{"reader":"grok-4.3"},"one_line_summary":"OpenAI releases two open-weight reasoning models, gpt-oss-120b and gpt-oss-20b, trained via distillation and RL with claimed strong results on math, coding, and safety benchmarks.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Two open-weight models using mixture-of-expert architecture deliver strong results on math, coding, and safety benchmarks while supporting agentic tool use.","strongest_claim":"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.","weakest_assumption":"The assumption that the internal training process and benchmark evaluations accurately reflect real-world agentic performance and safety without detailed public evidence or specific scores provided in the abstract."}},"verdict_id":"0baf52b7-c120-45cb-844b-3e9013774d62"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:d0aa9311ad7646a406b8e819dc6dd85441a1b957b366ea7a0de54b42d285ccb9","target":"record","created_at":"2026-07-05T11:54:15Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"e163e6febcae44ec88546116a93ce4b987d40bc98f4d9887a12b9907100fbef1","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-08-08T19:24:38Z","title_canon_sha256":"bb7b42f5055e7b4fc5e5e22b75da7aae00025a373a573422058f2fd7342fce08"},"schema_version":"1.0","source":{"id":"2508.10925","kind":"arxiv","version":1}},"canonical_sha256":"18f49ac97066c4cd8fb0af5888c6136e058f6d471017344d133edadf03aa70d1","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"18f49ac97066c4cd8fb0af5888c6136e058f6d471017344d133edadf03aa70d1","first_computed_at":"2026-07-05T11:54:15.454049Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T11:54:15.454049Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"78t509B3BpNlLw4q/AIcDX6V8Mym8L9NI4D63nhRJhrAb9KiduPppjV3qzZjoM5e3edN3GA2ObM6Gj2d+p/aCA==","signature_status":"signed_v1","signed_at":"2026-07-05T11:54:15.454537Z","signed_message":"canonical_sha256_bytes"},"source_id":"2508.10925","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:d0aa9311ad7646a406b8e819dc6dd85441a1b957b366ea7a0de54b42d285ccb9","sha256:877d8b2f9919dc6d38f5f2864edd3c6a590a10c196683ec789043c4e1a9fe849"],"state_sha256":"a285b310c47a8387b53382e7a1847a6776f7a7eb774aa8b6963196734743643d"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"Y0OpRF0XCbPS2MXoNuxud1Qt0qCATguZboVkMwyDRb26ymGOCswHyp66ko24goQ3WhSJOukwlbyIyGrLcUDmDw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-04T09:17:20.344424Z","bundle_sha256":"97fa2ba22df9da1f639b71fa6a8bb4429498041663978eff6bc0aeeaa2757c7e"}}