Meta releases Muse Spark 1.3, achieving state-of-the-art benchmark scores and improved efficiency
Meta released Muse Spark 1.3, its latest AI model, via Muse Code and the Meta Model API. The model achieves a state-of-the-art score on the DeepSWE v1.1 benchmark, with 20% fewer tool calls and 25% fewer tokens than version 1.2. It scored 98.5 on the MRCR 256K–512K evaluation, beating GPT-5.6 Sol, and approaches Opus 5 on several benchmarks. Alexandr Wang claimed it outperforms OpenAI's GPT-5.6 Sol at coding.
Editorial responsibility
- No named human review is recorded for this page.
- Reports are grouped by semantic similarity and deterministic rules. Language models may assist titles, summaries, translation and cross-source analysis; the page itself is projected from evidence records.
- Current automated evidence projection
Cross-source coverage
Wire timeline
Meta Releases Muse Spark 1.3, Scores 62 on Intelligence Index Near Claude and GPT-5.6
Meta has released Muse Spark 1.3, the fourth version of its Muse Spark AI model in five months. According to real-world test data from Artificial Analysis, the max variant of Muse Spark 1.3, available as a limited-time partner preview, scored 62 on the Artificial Analysis Intelligence Index. This score places the model's capabilities and costs in a comparative framework alongside offerings from OpenAI, Anthropic, and xAI, approaching the performance levels of Claude and GPT-5.6. The release marks a rapid iteration cycle for Meta's AI model family.
Meta Releases Muse Spark 1.3 with Improved Agent and Scientific Reasoning
Meta has officially released Muse Spark 1.3, marking the fourth version of the AI model in five months. The update includes improvements in agent and scientific reasoning capabilities. According to an announcement by Scale AI founder and Meta Chief AI Officer Alexandr Wang on X, the 'xhigh' version of Muse Spark 1.3 scored 61 points on the Artificial Analysis Intelligence Index. Meta has also published complete evaluation data for this release, allowing readers to compare changes in agent performance and single-task cost. This rapid iteration cycle underscores Meta's aggressive pace in advancing its AI model lineup.
JUST IN: Meta releases Muse Spark 1.3, its most powerful AI model yet, with Alexandr Wang claiming it is “better than” OpenAI’s GPT-5.6 Sol at coding.
Meta has released Muse Spark 1.3, described as its most powerful AI model to date. Alexandr Wang, a notable figure in the AI industry, claimed that the new model is superior to OpenAI's GPT-5.6 Sol specifically in coding tasks. The announcement was made via a post on Polymarket, a prediction market platform. This release marks a significant step in the ongoing competition between major AI developers, with Meta directly challenging OpenAI's leading model in a key performance area. The claim, if substantiated, could shift perceptions of model capability in the rapidly evolving AI landscape.
Show 4 older updatesHide older updates
Muse Spark 1.3 AI model released on OpenRouter for agentic and coding workflows
Muse Spark 1.3 has been released on the OpenRouter platform. The new version is specifically built for long-running agentic, multi-agent, and coding workflows. Key features include the ability to track what it learns over time, handle conflicting inputs, and ask for clarification or confirmation when needed. The announcement was made via a post on X (formerly Twitter) with a link to access the model on OpenRouter. This release targets developers and users working on complex, extended AI-driven tasks that require persistent context and adaptive reasoning.
Meta releases Muse Spark 1.3 with 20% fewer tool calls and 25% fewer tokens
Meta has released Muse Spark 1.3, a new version of its AI model. According to reports, the model achieves 20% fewer tool calls and 25% fewer tokens compared to version 1.2, suggesting improved efficiency for coding tasks. The model also demonstrates strong performance on long-context benchmarks, scoring 98.5 on the MRCR 256K–512K evaluation, which reportedly beats GPT-5.6 Sol. Additionally, Muse Spark 1.3 approaches the performance of Opus 5 on several benchmarks, including JobBench (64.9 vs 65.7), OSWorld (66.9 vs 68.3), and AutomationBench (49.4 vs 50.3). The release signals continued advancement in AI model efficiency and long-context understanding.
Meta releases Muse Spark 1.3 AI model via Muse Code and Meta Model API
Meta has released Muse Spark 1.3, the latest version of its AI model, which is rolling out today in Muse Code and the Meta Model API. The announcement follows a rapid sequence of AI releases over the past 30 hours, including Fable 5.1 and Gemini 3.8 Flash. The post indicates that a 'max reasoning' feature for Muse Spark 1.3 is coming soon. The release underscores the accelerating pace of AI model updates from major technology companies.
Muse Spark 1.3 achieves state-of-the-art score on DeepSWE v1.1 benchmark
Muse Spark 1.3 has been released, according to a post on X by synthwavedd. The new version of the AI model achieves a state-of-the-art (SoTA) score on the DeepSWE v1.1 benchmark, indicating strong performance. The post notes that this is part of a very busy week for releases or developments, though no further details are provided. The benchmark result suggests Muse Spark 1.3 outperforms previous models on this specific evaluation, which is likely focused on software engineering or coding tasks given the benchmark name. The announcement is brief but confirms a concrete milestone for the model's development.