{"event":{"id":"evt-whisper-open-20220921","dedupe_key":"open_speech_model_release:2022-09-21:openai-releases-whisper-models-and-inference-code","event_type":"open_speech_model_release","title":"OpenAI releases Whisper models and inference code","summary":"OpenAI open-sourced Whisper models and inference code for multilingual transcription and translation after training on 680,000 hours of multilingual and multitask supervised web data.","occurred_at":"2022-09-21T00:00:00.000Z","published_at":"2022-09-21T00:00:00.000Z","observed_at":"2026-10-02T15:12:29.000Z","reconstructed_at":"2026-10-02T15:12:29.000Z","ingest_type":"live_world_scan","locations":null,"status":"verified","confidence":0.99,"metadata":{"actors":["OpenAI"],"what_happened_at_time":"A multilingual speech-recognition and translation model family became directly reusable through released models and inference code.","evidence_at_time":"OpenAI's September 21 release documents the training-data scale, multilingual tasks and open-source release.","affected_layers":["layer-models","layer-open-source","layer-software","layer-human","layer-ai"],"change_kind":"open_multimodal_model_distribution","q2_continuity":"Q2 multimodal work was largely visual/generative; Q3 broadened open model distribution into speech.","retrospective_inference":"Reusable model artifacts were spreading across modalities rather than remaining confined to text or image interfaces.","uncertainty":"OpenAI's claim of near-human robustness is vendor-reported and is not required for this reconstruction."},"created_at":"2026-10-02T16:08:16.247Z","updated_at":"2026-10-02T16:08:16.247Z"},"artifacts":[{"id":"art-whisper-20220921","canonical_url":"https://openai.com/index/whisper/","artifact_type":"open_model_release","title":"Whisper open-source speech model","summary":"OpenAI released Whisper models and inference code for multilingual speech recognition and translation, trained on 680,000 hours of multilingual and multitask supervised web data.","creator_entities":["OpenAI"],"released_at":"2022-09-21T00:00:00.000Z","content_hash":null,"metadata":{"historical_scope":"2022-Q3"},"created_at":"2026-10-02T16:08:16.239Z","updated_at":"2026-10-02T16:08:16.239Z","relation":"evidenced_by"}],"signals":[{"id":"sig-q3-open-model-distribution","statement":"Model access broadened materially from controlled frontier research access toward downloadable or openly reusable model artifacts across language, translation, image generation and speech, while licenses and hardware requirements remained real constraints.","signal_type":"model_access_and_distribution_shift","direction":"increasing","confidence":0.97,"epistemic_status":"supported_inference","mapping_mode":"retrospective","reconstructed_at":"2026-10-02T15:12:29.000Z","created_at":"2026-10-02T16:08:16.249Z","updated_at":"2026-10-02T16:08:16.249Z"}]}