{ "version": 1, "event_id": "evt_294b831ec4a5f79c", "url": "https://xiyu.news/events/evt_294b831ec4a5f79c/", "json": "https://xiyu.news/api/events/evt_294b831ec4a5f79c.json", "type": "product_release", "status": "monitoring", "category": "technology", "title": { "zh": "Meta发布Muse Wonder与Muse Wonder Full前沿AI模型", "en": "Meta releases Muse Wonder and Muse Wonder Full frontier AI models" }, "current_state": { "zh": "Meta已于2026年10月12日发布Muse Wonder与Muse Wonder Full，二者具备100万token上下文窗口，并引入Active Memory与Active Focus能力；Muse Wonder通过Meta Model API与Muse Code上线，标准版每百万输入token收费4美元，Muse Wonder Full仅面向Muse Code每月50美元或200美元订阅档。Meta自称迈向“个人超级智能”，自评基准领先GPT-6 Astra与Claude Opus 5.5；但Artificial Analysis独立评分显示Wonder Full为56、Wonder为52，高于Fable 5.1与Astra的53，仍低于Opus 5.5的58，且Abacus.AI的Bindu Reddy质疑其性能。", "en": "Meta released Muse Wonder and Muse Wonder Full on 2026-10-12 with a 1 million token context window and new Active Memory and Active Focus capabilities; Muse Wonder is live via Meta Model API and Muse Code, with standard Wonder charging $4 per million input tokens, while Muse Wonder Full is limited to Muse Code's $50 or $200 monthly subscription tiers. Meta calls it a step toward 'personal superintelligence' and its own benchmarks put the models ahead of GPT-6 Astra and Claude Opus 5.5, but independent Artificial Analysis scores show Wonder Full at 56 and Wonder at 52, above Fable 5.1 and Astra at 53 yet below Opus 5.5 at 58, with Abacus.AI's Bindu Reddy challenging the performance claims." }, "first_seen_at": "2026-10-12T20:21:34.287288+00:00", "last_updated_at": "2026-10-13T03:47:41.540020+00:00", "last_material_change_at": "2026-10-13T03:47:41.540020+00:00", "confidence": 0.75, "updates_count": 3, "sources_count": 5, "entities": [ "ai-models", "anthropic", "finally", "frontier", "frontier-model", "full", "meta", "muse", "new", "openai", "puts", "race", "wonder", "zuckerberg" ], "identifiers": [ "gpt-6" ], "topics": [ "ai-competition", "ai-labs", "ai-models", "benchmarks", "coding", "frontier-ai", "frontier-model", "frontier-models", "llm-benchmarks", "llm-release", "meta", "model-release", "superintelligence-labs" ], "updates": [ { "update_id": "upd_32a6563662504ddd", "event_id": "evt_294b831ec4a5f79c", "occurred_at": "2026-10-12T17:00:00Z", "published_at": "2026-10-12T17:00:00Z", "first_seen_at": "2026-10-12T20:21:34.287288Z", "time_precision": "published", "update_type": "initial", "material_change": true, "title_zh": "Meta 推出了几款 Wonder Full 新 AI 模型", "title_en": "Meta has some Wonder Full new AI models", "what_changed_zh": "Meta 从其超级智能实验室推出了两款新 AI 模型 Muse Wonder 和 Muse Wonder Full，声称它们能与业界领先的模型竞争。", "what_changed_en": "Meta launched two new AI models, Muse Wonder and Muse Wonder Full, from its Superintelligence Labs, claiming they can compete with the industry's leading models.", "current_state_zh": "Meta 从其超级智能实验室推出了两款新 AI 模型 Muse Wonder 和 Muse Wonder Full，声称它们能与业界领先的模型竞争。", "current_state_en": "Meta launched two new AI models, Muse Wonder and Muse Wonder Full, from its Superintelligence Labs, claiming they can compete with the industry's leading models.", "detailed_summary_zh": "Meta launched two new AI models, Muse Wonder and Muse Wonder Full, from its Superintelligence Labs, claiming they can compete with the industry's leading models.", "detailed_summary_en": "Meta launched two new AI models, Muse Wonder and Muse Wonder Full, from its Superintelligence Labs, claiming they can compete with the industry's leading models.", "background_zh": "", "background_en": "", "community_discussion_zh": "", "community_discussion_en": "", "market_impact_zh": "", "market_impact_en": "", "importance_score": 8.0, "references": [], "confidence": 0.75, "story_ids": [ "rss:www.theverge.com_rss_ai-artificial-intelligence_index.xml:7ebc7763a031c860", "telegram:theblockbeats:200115", "rss:techcrunch.com_category_artificial-intelligence_feed_:0c806980bc61417f" ], "sources": [ { "url": "https://www.theverge.com/ai-artificial-intelligence/1009479/meta-muse-wonder-full-ai-model", "label": "The Verge AI", "source_type": "rss", "official": false }, { "url": "https://m.theblockbeats.info/flash/371449?from=telegram", "label": "theblockbeats", "source_type": "telegram", "official": false }, { "url": "https://techcrunch.com/2026/10/12/meta-releases-its-first-frontier-model-muse-wonder/", "label": "TechCrunch AI", "source_type": "rss", "official": false } ] }, { "update_id": "upd_8a17ea4a6c8a4312", "event_id": "evt_294b831ec4a5f79c", "occurred_at": "2026-10-12T20:46:04Z", "published_at": "2026-10-12T20:46:04Z", "first_seen_at": "2026-10-12T21:21:47.186820Z", "time_precision": "published", "update_type": "confirmation", "material_change": true, "title_zh": "Meta 的新 Muse Wonder 终于让扎克伯格跻身前沿 AI 竞赛", "title_en": "Meta's New Muse Wonder Finally Puts Zuckerberg in the Frontier AI Race", "what_changed_zh": "新报道补充了Muse Wonder与Muse Wonder Full发布的具体细节：100万token上下文窗口、API版Muse Wonder定价为每百万输入/输出token 4美元/20美元、通过Meta Model API与Muse Code上线（Full仅供付费Muse Code订阅者），并给出Meta自评基准（Personal Superintelligence Index 59.1、SWE-Marathon 57.5、Harvey Legal Agent 13.1）及Artificial Analysis排名仍将Opus 5.5列第一。", "what_changed_en": "New reporting adds specifics of the Muse Wonder / Muse Wonder Full launch: 1 million token context window, API version Muse Wonder priced at $4/$20 per million input/output tokens, availability now via Meta Model API and Muse Code (Full reserved for paying Muse Code subscribers), Meta's own benchmark scores (Personal Superintelligence Index 59.1, SWE-Marathon 57.5, Harvey Legal Agent 13.1), and Artificial Analysis still ranking Opus 5.5 first.", "current_state_zh": "Meta已于2026年10月12日发布Muse Wonder与Muse Wonder Full，二者具备100万token上下文窗口，Muse Wonder已通过Meta Model API与Muse Code上线（Muse Wonder Full仅面向付费Muse Code订阅者），API定价为每百万输入/输出token 4美元/20美元；Meta自评基准领先GPT-6 Astra与Claude Opus 5.5，但第三方Artificial Analysis仍将Opus 5.5列为第一。", "current_state_en": "Meta released Muse Wonder and Muse Wonder Full on 2026-10-12 with a 1 million token context window; Muse Wonder is live via the Meta Model API and Muse Code (Muse Wonder Full reserved for paying Muse Code subscribers) at $4/$20 per million input/output tokens. Meta's own benchmarks put them ahead of GPT-6 Astra and Claude Opus 5.5, though third-party Artificial Analysis still ranks Opus 5.5 first.", "detailed_summary_zh": "Meta released Muse Wonder and Muse Wonder Full, the first frontier-class models from Meta Superintelligence Labs, featuring a 1 million token context window and pricing at $4/$20 per million input/output tokens, with Meta's own benchmarks placing them above GPT-6 Astra and Claude Opus 5.5 while Artificial Analysis still ranks Opus 5.5 first.", "detailed_summary_en": "Meta released Muse Wonder and Muse Wonder Full, the first frontier-class models from Meta Superintelligence Labs, featuring a 1 million token context window and pricing at $4/$20 per million input/output tokens, with Meta's own benchmarks placing them above GPT-6 Astra and Claude Opus 5.5 while Artificial Analysis still ranks Opus 5.5 first.", "background_zh": "Meta 在开源模型 Llama 4 于独立测试中表现不佳后，于 2025 年成立了 Meta Superintelligence Labs，此前其以约 140 亿美元收购 Scale AI 的股份，并由该公司创始人 Alexandr Wang 领导这一新部门。该部门的首个模型 Muse Spark 于 2026 年早些时候发布，是一款小而快的闭源模型。Muse Wonder 的问世距 OpenAI 于 9 月 3 日发布 GPT-6 Astra、Anthropic 于 9 月 22 日发布 Claude Opus 5.5 仅数周。", "background_en": "Meta formed Meta Superintelligence Labs in 2025 after its open-source Llama 4 models stumbled in independent tests, following a roughly $14 billion deal for Scale AI that put founder Alexandr Wang in charge of the unit. The group's first model, Muse Spark, launched earlier in 2026 as a small, fast, closed model. Muse Wonder arrives weeks after OpenAI shipped GPT-6 Astra on September 3 and Anthropic followed on September 22 with Claude Opus 5.5.", "community_discussion_zh": "", "community_discussion_en": "", "market_impact_zh": "直接传导渠道是对前沿推理定价的压力：Meta 以每百万 token 4 美元/20 美元与 Claude Opus 5.5 持平，并低于 GPT-6 Astra 的 10 美元/50 美元，这影响的是竞争实验室的 API 收入经济性，而非直接涉及任何加密资产。对加密市场的传导则经由市场情绪——Meta 跻身前沿实验室的定位会强化更广泛的 AI 资本支出与风险偏好叙事，这类叙事通常会让高贝塔的科技相关代币与其他风险资产一同波动。", "market_impact_en": "The direct channel is pricing pressure on frontier inference: Meta is matching Claude Opus 5.5 at $4/$20 per million tokens and undercutting GPT-6 Astra's $10/$50, which affects the API revenue economics of competing labs rather than any crypto asset directly. The read-through to crypto markets runs through sentiment — Meta's positioning as a frontier lab feeds the broader AI capex and risk-appetite narrative that typically moves high-beta technology-linked tokens alongside other risk assets.", "importance_score": 8.0, "references": [ { "url": "https://en.wikipedia.org/wiki/Meta_Superintelligence_Labs", "title": "Meta Superintelligence Labs" }, { "url": "https://artificialanalysis.ai/leaderboards/models", "title": "LLM Leaderboard - Comparison of AI models from... | Artificial Analysis" }, { "url": "https://en.wikipedia.org/wiki/Context_window", "title": "Context window - Wikipedia" } ], "confidence": 0.93, "story_ids": [ "rss:decrypt.co_feed:8c7aae0110c4420c" ], "sources": [ { "url": "https://decrypt.co/380768/meta-muse-wonder-zuckerberg-frontier-ai-race", "label": "Decrypt", "source_type": "rss", "official": false } ] }, { "update_id": "upd_f44635fcf1dbbe8c", "event_id": "evt_294b831ec4a5f79c", "occurred_at": "2026-10-13T01:54:31Z", "published_at": "2026-10-13T01:54:31Z", "first_seen_at": "2026-10-13T03:47:41.540020Z", "time_precision": "published", "update_type": "confirmation", "material_change": true, "title_zh": "Meta 称新的 Muse Wonder AI 模型可与 OpenAI、Anthropic 匹敌", "title_en": "Meta claims new Muse Wonder AI models rival OpenAI, Anthropic", "what_changed_zh": "新报道补充了Muse Wonder与Muse Wonder Full发布后的更多细节：两款模型引入Active Memory与Active Focus能力；Artificial Analysis独立评分显示Muse Wonder Full为56、Muse Wonder为52，高于Claude Fable 5.1与GPT-6 Astra的53，但低于Claude Opus 5.5的58；Abacus.AI的Bindu Reddy质疑Muse Wonder表现；标准版Wonder每百万输入token收费4美元，Wonder Full仅限Muse Code每月50美元或200美元订阅档。", "what_changed_en": "New reporting adds more post-launch details for Muse Wonder and Muse Wonder Full: the models introduce Active Memory and Active Focus capabilities; Artificial Analysis independent scores show Muse Wonder Full at 56 and Muse Wonder at 52, above Claude Fable 5.1 and GPT-6 Astra at 53 but below Claude Opus 5.5 at 58; Abacus.AI's Bindu Reddy challenged Muse Wonder's performance; standard Wonder charges $4 per million input tokens, and Wonder Full is limited to Muse Code's $50 or $200 monthly subscription tiers.", "current_state_zh": "Meta已于2026年10月12日发布Muse Wonder与Muse Wonder Full，二者具备100万token上下文窗口，并引入Active Memory与Active Focus能力；Muse Wonder通过Meta Model API与Muse Code上线，标准版每百万输入token收费4美元，Muse Wonder Full仅面向Muse Code每月50美元或200美元订阅档。Meta自称迈向“个人超级智能”，自评基准领先GPT-6 Astra与Claude Opus 5.5；但Artificial Analysis独立评分显示Wonder Full为56、Wonder为52，高于Fable 5.1与Astra的53，仍低于Opus 5.5的58，且Abacus.AI的Bindu Reddy质疑其性能。", "current_state_en": "Meta released Muse Wonder and Muse Wonder Full on 2026-10-12 with a 1 million token context window and new Active Memory and Active Focus capabilities; Muse Wonder is live via Meta Model API and Muse Code, with standard Wonder charging $4 per million input tokens, while Muse Wonder Full is limited to Muse Code's $50 or $200 monthly subscription tiers. Meta calls it a step toward 'personal superintelligence' and its own benchmarks put the models ahead of GPT-6 Astra and Claude Opus 5.5, but independent Artificial Analysis scores show Wonder Full at 56 and Wonder at 52, above Fable 5.1 and Astra at 53 yet below Opus 5.5 at 58, with Abacus.AI's Bindu Reddy challenging the performance claims.", "detailed_summary_zh": "Meta unveiled the Muse Wonder and Muse Wonder Full AI models, claiming they rival OpenAI and Anthropic offerings and marking a step toward 'personal superintelligence,' with independent scores placing the top model near but not at the head of the frontier leaderboard.", "detailed_summary_en": "Meta unveiled the Muse Wonder and Muse Wonder Full AI models, claiming they rival OpenAI and Anthropic offerings and marking a step toward 'personal superintelligence,' with independent scores placing the top model near but not at the head of the frontier leaderboard.", "background_zh": "", "background_en": "", "community_discussion_zh": "", "community_discussion_en": "", "market_impact_zh": "", "market_impact_en": "", "importance_score": 7.5, "references": [], "confidence": 0.93, "story_ids": [ "rss:cointelegraph.com_rss:5ab60e4632f0a3d0" ], "sources": [ { "url": "https://cointelegraph.com/news/meta-launches-muse-wonder-ai-models-challenging-top-rivals?utm_source=rss_feed&utm_medium=rss&utm_campaign=rss_partner_inbound", "label": "Cointelegraph", "source_type": "rss", "official": false } ] } ] }