--- title: Thinking Machines Lab drops Inkling & Meta’s Muse Spark 1.1 description: A podcast episode covering four major AI stories: Thinking Machines Lab's first model release 'Inkling' (975B parameter MoE, open-weight, natively multimodal... canonical: https://daily.dev/posts/thinking-machines-lab-drops-inkling-meta-s-muse-spark-1-1-adpqmjkzj twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Thinking Machines Lab drops Inkling & Meta’s Muse Spark 1.1 | daily.dev og:description: A podcast episode covering four major AI stories: Thinking Machines Lab's first model release 'Inkling' (975B parameter MoE, open-weight, natively multimodal... og:url: https://daily.dev/posts/thinking-machines-lab-drops-inkling-meta-s-muse-spark-1-1-adpqmjkzj og:image: https://api.daily.dev/og/posts/adpqMjKzJ.png og:image:alt: Thinking Machines Lab drops Inkling & Meta’s Muse Spark 1.1 og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Thinking Machines Lab drops Inkling & Meta’s Muse Spark 1.1 **[IBM Technology](https://daily.dev/sources/ibmtechnology)** · 39 min read · 3 upvotes · 0 comments ## Summary A podcast episode covering four major AI stories: Thinking Machines Lab's first model release 'Inkling' (975B parameter MoE, open-weight, natively multimodal with a self-improving fine-tuning platform called Tinker); Meta's Muse Spark 1.1 launch (closed model targeting agent workloads, cost efficiency, and enterprise use cases); OpenAI's GPT 5.6 Soul scoring 8% on ARC-AGI 3 benchmark and what that means for AGI progress; and Anthropic's controversial paper on Claude's internal 'JSP space' suggesting a subconscious-like processing layer, with debate on its implications for AI safety, agent monitoring, and interpretability. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### How many parameters does Thinking Machines Lab's Inkling model have and is it open source? Inkling is a 975-billion total parameter mixture-of-experts model with 41 billion active parameters, released as an open-weight model by Thinking Machines Lab, the startup founded by former OpenAI CTO Mira Murati. It supports a 1 million token context window, true multimodal token-level processing, and pairs with the Tinker fine-tuning API for automated post-training loops. _Developers evaluating open-weight models with fine-tuning platforms can track releases like this on daily.dev._ ### What score did OpenAI's GPT-5.6 Solus achieve on the ARC-AGI 3 benchmark? GPT-5.6 Solus scored around 8% on ARC-AGI 3, up from roughly 0% for prior models like Luna and Terra on the same test. The jump is considered notable given the benchmark's difficulty, though panelists note a 12-year-old human can score close to 100%, so it falls well short of demonstrating general intelligence. _Anyone tracking AGI benchmark progress can follow model score updates like this on daily.dev._ ### What is the Jacobian space Anthropic describes in its Claude interpretability research? The Jacobian space, or JSpace, refers to internal regions of Claude's neural network where reasoning and processing occur before being projected into token outputs, analogous to subconscious computation. Anthropic uses a technique based on averaging activations rather than projecting into token space, and Google DeepMind's Neel Nanda independently replicated some results on open-source models. _Engineers building on interpretability findings for AI safety can follow research breakdowns like this on daily.dev._ ## Similar posts on daily.dev - [Meta enters the crowded AI coding battle with Muse Spark 1.1](https://daily.dev/posts/meta-enters-the-crowded-ai-coding-battle-with-muse-spark-1-1-wtfjrleyt) · TechCrunch · 1 upvotes · 0 comments - [Meta’s Muse Spark: a smaller, faster AI model for broad app deployment](https://daily.dev/posts/meta-s-muse-spark-a-smaller-faster-ai-model-for-broad-app-deployment-rxhf7mbsh) · InfoWorld · 2 upvotes · 0 comments --- Tags: [#ai](https://daily.dev/tags/ai), [#open-source](https://daily.dev/tags/open-source), [#llm](https://daily.dev/tags/llm), [#ai-agents](https://daily.dev/tags/ai-agents), [#deep-learning](https://daily.dev/tags/deep-learning) [View this post on daily.dev](https://daily.dev/posts/thinking-machines-lab-drops-inkling-meta-s-muse-spark-1-1-adpqmjkzj) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Thinking Machines Lab drops Inkling & Meta’s Muse Spark 1.1","url":"https://daily.dev/posts/thinking-machines-lab-drops-inkling-meta-s-muse-spark-1-1-adpqmjkzj","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/thinking-machines-lab-drops-inkling-meta-s-muse-spark-1-1-adpqmjkzj"},"datePublished":"2026-07-17T10:27:31.341Z","dateModified":"2026-10-15T02:22:10.825Z","description":"A podcast episode covering four major AI stories: Thinking Machines Lab's first model release 'Inkling' (975B parameter MoE, open-weight, natively multimodal...","image":"https://i.ytimg.com/vi/8rGYGFmytQs/sddefault.jpg","thumbnailUrl":"https://i.ytimg.com/vi/8rGYGFmytQs/sddefault.jpg","isAccessibleForFree":true,"articleSection":"IBM Technology","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"IBM Technology","logo":"https://media.daily.dev/image/upload/s--OXHaOiWX--/f_auto/v1715498011/logos/ibmtechnology","url":"https://daily.dev/sources/ibmtechnology"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/thinking-machines-lab-drops-inkling-meta-s-muse-spark-1-1-adpqmjkzj","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":3},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai,open-source,llm,ai-agents,deep-learning","timeRequired":"PT39M","video":{"@type":"VideoObject","name":"Thinking Machines Lab drops Inkling & Meta’s Muse Spark 1.1","description":"A podcast episode covering four major AI stories: Thinking Machines Lab's first model release 'Inkling' (975B parameter MoE, open-weight, natively multimodal...","thumbnailUrl":"https://i.ytimg.com/vi/8rGYGFmytQs/sddefault.jpg","uploadDate":"2026-07-17T10:27:31.341Z","duration":"PT39M","url":"https://api.daily.dev/r/adpqMjKzJ","embedUrl":"https://www.youtube.com/embed/8rGYGFmytQs"}} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"IBM Technology","item":"https://daily.dev/sources/ibmtechnology"},{"@type":"ListItem","position":3,"name":"Thinking Machines Lab drops Inkling & Meta’s Muse Spark 1.1"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/thinking-machines-lab-drops-inkling-meta-s-muse-spark-1-1-adpqmjkzj#faq","mainEntity":[{"@type":"Question","name":"How many parameters does Thinking Machines Lab's Inkling model have and is it open source?","acceptedAnswer":{"@type":"Answer","text":"Inkling is a 975-billion total parameter mixture-of-experts model with 41 billion active parameters, released as an open-weight model by Thinking Machines Lab, the startup founded by former OpenAI CTO Mira Murati. It supports a 1 million token context window, true multimodal token-level processing, and pairs with the Tinker fine-tuning API for automated post-training loops. Developers evaluating open-weight models with fine-tuning platforms can track releases like this on daily.dev."}},{"@type":"Question","name":"What score did OpenAI's GPT-5.6 Solus achieve on the ARC-AGI 3 benchmark?","acceptedAnswer":{"@type":"Answer","text":"GPT-5.6 Solus scored around 8% on ARC-AGI 3, up from roughly 0% for prior models like Luna and Terra on the same test. The jump is considered notable given the benchmark's difficulty, though panelists note a 12-year-old human can score close to 100%, so it falls well short of demonstrating general intelligence. Anyone tracking AGI benchmark progress can follow model score updates like this on daily.dev."}},{"@type":"Question","name":"What is the Jacobian space Anthropic describes in its Claude interpretability research?","acceptedAnswer":{"@type":"Answer","text":"The Jacobian space, or JSpace, refers to internal regions of Claude's neural network where reasoning and processing occur before being projected into token outputs, analogous to subconscious computation. Anthropic uses a technique based on averaging activations rather than projecting into token space, and Google DeepMind's Neel Nanda independently replicated some results on open-source models. Engineers building on interpretability findings for AI safety can follow research breakdowns like this on daily.dev."}}]} ```