{"rewrite":{"id":"r_cfb9858270f670b1719835b7","clusterId":"c_ac956268b860597eb2d437bf","slug":"meta-announces-muse-wonder-ai-models-claiming-benchmark-wins-over-openai-and-anthropic","model":"deepseek-v4-1-flash","headline":"Meta Announces Muse Wonder AI Models Claiming Benchmark Wins Over OpenAI and Anthropic","summary":"Meta announced two frontier AI models, Muse Wonder and Muse Wonder Full, on October 12, 2026. The company presented benchmark results that it says surpass OpenAI's GPT-6 Astra and Anthropic's Claude Opus 5.5 on several tests. On DeepSWE 1.1, which measures long-term software development capability, Muse Wonder Full scored 79.0 and Muse Wonder scored 78.8, above GPT-6 Astra at 74.1 and Claude Opus 5.5 at 74.2. On Harvey LAB-AA, a legal document task, Full scored 13.1 against GPT-6 Astra's 8.6. The results are not uniform: on the healthcare benchmark HealthBench Pro, Full scored 62.3, below GPT-6 Astra's 65.3 and Claude Opus 5.5's 65.6. Both models share two features, Active Memory for retaining information across conversations and Active Focus for working through difficult tasks. Muse Wonder is available through the Meta Model API and Muse Code, while Muse Wonder Full is limited to upper-tier Muse Code plans.","whyItMatters":"Meta's benchmark claims put a third company's numbers on the table next to OpenAI and Anthropic, but the HealthBench Pro shortfall shows the lead is benchmark-specific, not across the board.","webCardHtml":"\u003cp\u003eMeta described Muse Wonder as a model that delivers frontier-level performance at high speed, and Muse Wonder Full as the full version of the model. On the Artificial Analysis Intelligence Index, which aggregates results across benchmark tests, Muse Wonder scored 52 and Muse Wonder Full scored 56.\u003c/p\u003e\n\u003cp\u003eMeta cites active memory and active focus as the key elements for making the models useful as personal and business agents. Active memory records important items about the user and searches for detailed information relevant to the task. Active focus continues work on difficult tasks by exploring alternatives and adapting to new information.\u003c/p\u003e\n\u003cp\u003eOn safety metrics, Meta presented graphs showing both the rate of executing unauthorized code and the rate of mishandling personal information lower than Anthropic\u0026#39;s and OpenAI\u0026#39;s models. Muse Wonder Full is not offered through the API. Muse Wonder API pricing is $4 per 1 million input tokens, $0.20 for cached input, and $20.00 for output.\u003c/p\u003e\n\u003cp\u003eMeta says the models were trained over 14 months at its gigawatt-class data center Prometheus, following a revamp of its AI development setup.\u003c/p\u003e","blueskyPost":"HealthBench Pro splits Meta's claims: Muse Wonder Full scored 62.3 against GPT-6 Astra's 65.3 and Claude Opus 5.5's 65.6. The benchmark wins Meta chose to publish are the ones it won.","twitterPost":"Meta's Muse Wonder leads DeepSWE 1.1 but trails GPT-6 Astra on HealthBench Pro, 62.3 to 65.3.","threadsPost":"Meta's Muse Wonder Full leads DeepSWE 1.1 at 79.0 and Harvey LAB-AA at 13.1, but drops to 62.3 on HealthBench Pro against GPT-6 Astra's 65.3. The benchmark wins are real, and so is the gap on healthcare tasks.","newsletterBlurb":"Meta introduced Muse Wonder and Muse Wonder Full, two frontier models it says beat OpenAI's and Anthropic's flagships on coding and legal benchmarks. The results hold on some tests and not others, with HealthBench Pro going the other way. Muse Wonder is out through the Meta Model API and Muse Code; the Full version sits behind upper-tier Muse Code plans.","attributionJson":"[{\"source\":\"ASCII.jp\",\"url\":\"https://ascii.jp/elem/000/004/441/4441895/?rss\",\"title\":\"Meta Suddenly Enters Frontier AI Model Race, With Benchmarks Beating Opus 5.5 and Astra\"},{\"source\":\"GIGAZINE\",\"url\":\"https://gigazine.net/news/20261013-meta-ai-muse-wonder/\",\"title\":\"Meta announces high-performance AI \\\"Muse Wonder\\\" that surpasses cutting-edge models from OpenAI and Anthropic, catching up to the frontier within six months of the Muse series' debut\"}]","lintFlagsJson":null,"lintHits":0,"costUsd":0,"inputTokens":12072,"outputTokens":1281,"status":"published","repairAttempts":0,"nextRepairAt":null,"factsAttemptedAt":1791865325,"createdAt":"2026-10-13T04:16:41.000Z","publishedAt":"2026-10-13T04:21:05.000Z","updatedAt":"2026-10-13T04:21:05.000Z"},"cluster":{"id":"c_ac956268b860597eb2d437bf","canonicalTitle":"OpenAIやAnthropicの最先端モデルを上回る高性能AI「Muse Wonder」をMetaが発表、Museシリーズ登場から半年で最先端に追いつく","representativeArticleId":"a_6cfa3b2ef1c0fba5b05e55fa","sourceCount":2,"writtenSourceCount":2,"writeAttempts":0,"isSolo":true,"entitiesJson":"{\"anime_titles\":[],\"manga_titles\":[],\"work_titles\":[],\"studios\":[],\"people\":[],\"type\":\"announcement\",\"domain\":\"other\",\"is_roundup\":false}","contentType":"news","status":"published","firstSeenAt":"2026-10-13T02:00:00.000Z","lastSeenAt":"2026-10-13T02:25:00.000Z","updatedAt":"2026-10-13T04:21:05.000Z"},"attribution":[{"source":"ASCII.jp","url":"https://ascii.jp/elem/000/004/441/4441895/?rss","title":"Meta、いきなりフロンティアAIモデル競争へ　Opus 5.5やAstraを上回るベンチマークも"},{"source":"GIGAZINE","url":"https://gigazine.net/news/20261013-meta-ai-muse-wonder/","title":"OpenAIやAnthropicの最先端モデルを上回る高性能AI「Muse Wonder」をMetaが発表、Museシリーズ登場から半年で最先端に追いつく"}],"entities":{"anime_titles":[],"manga_titles":[],"work_titles":[],"studios":[],"people":[],"type":"announcement","domain":"other","is_roundup":false},"keyFacts":["Meta announced the AI models Muse Wonder and Muse Wonder Full on October 12, 2026.","On the DeepSWE 1.1 benchmark, Muse Wonder Full scored 79.0 and Muse Wonder scored 78.8, above GPT-6 Astra at 74.1 and Claude Opus 5.5 at 74.2.","On HealthBench Pro, Muse Wonder Full scored 62.3, below GPT-6 Astra at 65.3 and Claude Opus 5.5 at 65.6.","Muse Wonder is available through the Meta Model API and Muse Code, while Muse Wonder Full is restricted to upper-tier Muse Code plans.","Meta lists API pricing for Muse Wonder at $4 per million input tokens, $0.20 for cached input, and $20.00 for output."]}