{"passport":{"slug":"foresight-v3","display_name":"Foresight V3","model_type":"language","parameter_count":"~117B (inferred from base model gpt-oss-120b; unconfirmed for fine-tune)","context_window":"128K (inferred from base model gpt-oss-120b; unconfirmed for fine-tune)","claimed_developer":"Foresight (Unknown Company/Team)","confirmed_developer":null,"developer_confirmed_at":null,"developer_source":null,"first_appeared_at":"2026-04-05T18:31:00.083702+00:00","first_appeared_on":"reddit_localllama","status":"confirmed","availability":null,"deployment_status":"not_deployed","availability_scope":null,"gate":1,"eu_ai_act":"GPAI","regulatory_status":{"eu_ai_act":"GPAI"},"open_source":null,"weights_available":null,"license":null,"provenance_chain":[],"parent_model":null,"model_family":null,"superseded_by":null,"supersedes":[],"recommended_replacement":null,"product_links":null},"events":[{"event_type":"first_appearance","event_date":"2026-04-05T18:31:00.083702+00:00","title":"First recorded in AI news","detail":"Foresight V3, a fine-tuned derivative of the open-source gpt-oss-120b model, has achieved top performance on Prophet Arena, a live forecasting benchmark from UChicago's SIGMA Lab. The model was trained using an innovative auto-generated dataset created through a 'future-as-label' approach that leverages timestamps in public news sources to create labeled training data without human annotation. The model outperforms all frontier models including GPT-5.2, Gemini 3 Pro, and Claude Opus 4.5 on both Overall and Sports forecasting categories.","source_url":"https://reddit.com/r/LocalLLaMA/comments/1sbd0rc/how_we_turned_a_small_opensource_model_into_the/","source_name":"reddit_localllama"}],"delta":{"days_unattributed":95,"claimed_developer":"Foresight (Unknown Company/Team)","confirmed_developer":null,"match":false}}