{ "version": 1, "event_id": "evt_70346e36be141901", "url": "https://xiyu.news/events/evt_70346e36be141901/", "json": "https://xiyu.news/api/events/evt_70346e36be141901.json", "type": "product_release", "status": "resolved", "category": "technology", "title": { "zh": "MiniMax正式发布M3.1 Flash：每百万Token最低0.15美元，23日开源", "en": "MiniMax正式发布M3.1 Flash：每百万Token最低0.15美元，23日开源" }, "current_state": { "zh": "MiniMax M3.1 Flash 已完成发布并开源权重：4280亿参数 MoE、Agent 成本降70%、社区许可含收入限制，DeepSWE v1.1 得分73.7%。", "en": "MiniMax M3.1 Flash is fully released with open weights: 428B-parameter MoE, 70% lower agent-task cost, community license with revenue restrictions, and a 73.7% DeepSWE v1.1 score." }, "first_seen_at": "2026-10-19T14:11:31.526792+00:00", "last_updated_at": "2026-10-23T01:51:15.144941+00:00", "last_material_change_at": "2026-10-23T01:51:15.144941+00:00", "confidence": 0.75, "updates_count": 2, "sources_count": 3, "entities": [ "agent", "flash", "minimax", "token" ], "identifiers": [], "topics": [ "api-pricing", "benchmarks", "inference-cost", "llm", "minimax", "model-release", "moe", "open-weights" ], "updates": [ { "update_id": "upd_175d806407ad9aea", "event_id": "evt_70346e36be141901", "occurred_at": "2026-10-19T12:13:02Z", "published_at": "2026-10-19T12:13:02Z", "first_seen_at": "2026-10-19T14:11:31.526792Z", "time_precision": "published", "update_type": "initial", "material_change": true, "title_zh": "MiniMax正式发布M3.1 Flash：每百万Token最低0.15美元，23日开源", "title_en": "MiniMax正式发布M3.1 Flash：每百万Token最低0.15美元，23日开源", "what_changed_zh": "MiniMax正式发布M3.1 Flash，确认其就是匿名“Space Bunny”模型，公布基准测试成绩（Automation-Bench 62.2%、DeepSWE 73.7%、BabyVision 91.0%），API定价为每百万输入token 0.15美元、输出0.60美元，并计划于10月23日开放权重。", "what_changed_en": "MiniMax officially released M3.1 Flash, confirming it was the anonymous 'Space Bunny' model, publishing benchmarks (62.2% Automation-Bench, 73.7% DeepSWE, 91.0% BabyVision), setting API pricing at $0.15 per million input tokens and $0.60 output, and planning to open the weights on October 23.", "current_state_zh": "MiniMax正式发布M3.1 Flash，确认其就是匿名“Space Bunny”模型，公布基准测试成绩（Automation-Bench 62.2%、DeepSWE 73.7%、BabyVision 91.0%），API定价为每百万输入token 0.15美元、输出0.60美元，并计划于10月23日开放权重。", "current_state_en": "MiniMax officially released M3.1 Flash, confirming it was the anonymous 'Space Bunny' model, publishing benchmarks (62.2% Automation-Bench, 73.7% DeepSWE, 91.0% BabyVision), setting API pricing at $0.15 per million input tokens and $0.60 output, and planning to open the weights on October 23.", "detailed_summary_zh": "MiniMax officially released M3.1 Flash, confirming it was the anonymous 'Space Bunny' model, publishing benchmarks (62.2% Automation-Bench, 73.7% DeepSWE, 91.0% BabyVision), setting API pricing at $0.15 per million input tokens and $0.60 output, and planning to open the weights on October 23.", "detailed_summary_en": "MiniMax officially released M3.1 Flash, confirming it was the anonymous 'Space Bunny' model, publishing benchmarks (62.2% Automation-Bench, 73.7% DeepSWE, 91.0% BabyVision), setting API pricing at $0.15 per million input tokens and $0.60 output, and planning to open the weights on October 23.", "background_zh": "", "background_en": "", "community_discussion_zh": "", "community_discussion_en": "", "market_impact_zh": "", "market_impact_en": "", "importance_score": 7.0, "references": [ { "url": "https://news.ycombinator.com/item?id=50154540", "title": "Community discussion" } ], "confidence": 0.75, "story_ids": [ "telegram:theblockbeats:201242", "hackernews:story:50154540" ], "sources": [ { "url": "https://m.theblockbeats.info/flash/372587?from=telegram", "label": "theblockbeats", "source_type": "telegram", "official": false }, { "url": "https://www.minimax.io/blog/minimax-m31f", "label": "jtg", "source_type": "hackernews", "official": false } ] }, { "update_id": "upd_2495bc3bdc881cb1", "event_id": "evt_70346e36be141901", "occurred_at": "2026-10-23T01:38:03Z", "published_at": "2026-10-23T01:38:03Z", "first_seen_at": "2026-10-23T01:51:15.144941Z", "time_precision": "published", "update_type": "resolution", "material_change": true, "title_zh": "MiniMax M3.1 Flash开源：4280亿参数，Agent任务成本降70%", "title_en": "MiniMax M3.1 Flash开源：4280亿参数，Agent任务成本降70%", "what_changed_zh": "MiniMax 于10月23日按计划开源 M3.1 Flash 权重，已上架 Hugging Face 和魔搭；披露模型为4280亿总参数 MoE（每次激活约230亿），采用稀疏注意力、量化训练与 DSpark 推测解码，Agent 任务成本较上代降70%、生成速度近3倍，权重采用 MiniMax Community License（年收入超2000万美元需书面授权）。", "what_changed_en": "MiniMax released the open weights for M3.1 Flash on October 23 as planned, listed on Hugging Face and ModelScope; disclosed a 428B-total-parameter MoE (~23B active) with sparse attention, quantized training and DSpark speculative decoding, 70% lower agent-task cost and ~3x generation speed, under the MiniMax Community License (written authorization required above $20M annual revenue).", "current_state_zh": "MiniMax M3.1 Flash 已完成发布并开源权重：4280亿参数 MoE、Agent 成本降70%、社区许可含收入限制，DeepSWE v1.1 得分73.7%。", "current_state_en": "MiniMax M3.1 Flash is fully released with open weights: 428B-parameter MoE, 70% lower agent-task cost, community license with revenue restrictions, and a 73.7% DeepSWE v1.1 score.", "detailed_summary_zh": "MiniMax released open weights for M3.1 Flash, a 428B-parameter MoE model featuring sparse attention and speculative decoding that cuts agent-task costs by 70% and nearly triples generation speed, under a community license with revenue-based commercial restrictions.", "detailed_summary_en": "MiniMax released open weights for M3.1 Flash, a 428B-parameter MoE model featuring sparse attention and speculative decoding that cuts agent-task costs by 70% and nearly triples generation speed, under a community license with revenue-based commercial restrictions.", "background_zh": "", "background_en": "", "community_discussion_zh": "", "community_discussion_en": "", "market_impact_zh": "", "market_impact_en": "", "importance_score": 7.5, "references": [], "confidence": 0.9, "story_ids": [ "telegram:theblockbeats:201960" ], "sources": [ { "url": "https://m.theblockbeats.info/flash/373302?from=telegram", "label": "theblockbeats", "source_type": "telegram", "official": false } ] } ] }