{"schema":"postcutoff/event@1","as_of":"2026-11-22T13:32:00+01:00","url":"https://postcutoff.com/e/2026-09-02-meta-muse-spark-1-3/","md":"https://postcutoff.com/e/2026-09-02-meta-muse-spark-1-3/index.md","disclosure":{"written_by":"AI agents (Claude Opus 5.5 in Claude Code)","editor":"human editor","policy":"https://postcutoff.com/about/"},"license":null,"id":"2026-09-02-meta-muse-spark-1-3","date":"2026-09-02","date_precision":"day","short_title":"Meta releases Muse Spark 1.3 with max reasoning for long-horizon agentic work","deck":null,"takeaway":"On 2026-09-02 Meta Superintelligence Labs released Muse Spark 1.3 on Muse Code and the Meta Model API (muse-spark-1.3).","category":"model-release","category_label":"Model releases","importance":3,"confidence":"high","status":{"key":"confirmed","labels":["Confirmed"]},"sources":[{"n":1,"title":"Meta AI - Introducing Muse Spark 1.3","url":"https://research.meta.ai/blog/introducing-muse-spark-1-3","type":"official","group":"primary","domain":"research.meta.ai"},{"n":2,"title":"Meta - Muse Spark 1.3 multimodal evaluation methodology","url":"https://research.meta.ai/static/muse-spark-1-3-multimodal-evaluation-methodology","type":"paper","group":"primary","domain":"research.meta.ai"},{"n":3,"title":"Meta Model API - Muse Spark model page","url":"https://dev.meta.ai/models/muse-spark/","type":"docs","group":"primary","domain":"dev.meta.ai"},{"n":4,"title":"Meta for Developers on X","url":"https://x.com/MetaforDevs/status/2095232442953236714","type":"official","group":"primary","domain":"x.com"}],"official":4,"filed":"2026-09-30","updated":"2026-10-12","orgs":["Meta"],"title":"Meta releases Muse Spark 1.3 with max reasoning for long-horizon agentic work","summary":"On 2026-09-02 Meta Superintelligence Labs released Muse Spark 1.3 on Muse Code and the Meta Model API (`muse-spark-1.3`). It adds a max reasoning setting and is tuned for long, multi-workflow agentic threads. Meta says it uses ~20% fewer tool calls and ~25% fewer tokens than 1.2 on coding and is more robust to prompt injection. Meta repeated that Muse Spark open weights and bigger models are on its roadmap.","key_facts":["Released 2026-09-02 on Muse Code and the Meta Model API, with 'max reasoning' for hard reasoning and agentic tasks","Asks clarifying questions, asks the user for help when stuck, and confirms before consequential actions","Coding vs Muse Spark 1.2 (Meta engineers' comparisons): ~20% fewer tool calls, ~25% fewer tokens, less verbose","Safety: stronger resistance to adversarial inputs and prompt injection; better calibration on irreversible actions","API list price $1.25 input / $4.25 output per 1M tokens; 1M context, 200K max output; cheaper '-contributor' tier (see model file)","Roadmap per Meta: 'bigger models, the Muse Spark open weights release, and more'","Successor tier: on Oct 12, 2026 Meta released the larger Muse Wonder and Muse Wonder Full (entry 2026-10-12-meta-muse-wonder); Artificial Analysis scores them 52 and 56 on its Intelligence Index against 48 for Muse Spark 1.3 (max). Muse Spark 1.3 stays available at $1.25/$4.25"],"key_numbers":[],"tags":["meta","msl","muse","llm","agents","coding","api"],"science":null,"body_md":"## What happened\n\nMeta released Muse Spark 1.3 one month after 1.2. The update targets long agentic work: the model builds its own context from\nmessy, conflicting sources, keeps track of several workflows in one long thread, follows long instructions without dropping\nconstraints, and is trained across many harnesses. It is also trained to know better what it cannot do, so it reports blockers\ninstead of hallucinating results. It went live the same day in Muse Code and the Meta Model API.\n\n## Why it matters\n\nMuse Spark 1.3 is the model generation that shipped just before Meta's consumer Muse agent (Sept 8), and it is what developers can\ncall directly. Meta again promised open weights for Muse Spark but gave no date.\n\nThe pricing and context numbers come from Meta's developer pages (see data/models/muse-spark-1-3.md). Meta's blog gives no headline\nbenchmark numbers.","disputed":[],"related":[{"id":"2026-10-12-meta-muse-wonder","url":"https://postcutoff.com/e/2026-10-12-meta-muse-wonder/","date":"2026-10-12","date_precision":"day","short_title":"Meta releases Muse Wonder and Muse Wonder Full, its first frontier-class models","deck":null,"takeaway":"Meta had been absent from the top tier since the Llama line fell behind; Muse Spark (April 2026) was positioned as a lightweight model.","category":"model-release","category_label":"Model releases","importance":4,"confidence":"high","status":{"key":"confirmed","labels":["Confirmed"]},"sources":48,"official":10,"filed":"2026-10-12","updated":"2026-10-13","orgs":["Meta"]},{"id":"2026-09-08-meta-muse-personal-agent","url":"https://postcutoff.com/e/2026-09-08-meta-muse-personal-agent/","date":"2026-09-08","date_precision":"day","short_title":"Meta launches Muse, a free consumer personal AI agent","deck":null,"takeaway":"It is the first mass-market, free, always-on autonomous agent from a company with ~3.6 billion daily users, pushing agentic AI from developer tools into mainstream consumer use - with obvious safety and privacy stakes.","category":"agents","category_label":"Agents","importance":4,"confidence":"high","status":{"key":"confirmed","labels":["Confirmed"]},"sources":37,"official":4,"filed":"2026-09-29","updated":"2026-10-12","orgs":["Meta"]},{"id":"2026-08-05-meta-muse-code-spark-1-2","url":"https://postcutoff.com/e/2026-08-05-meta-muse-code-spark-1-2/","date":"2026-08-05","date_precision":"day","short_title":"Meta launches Muse Code terminal coding agent powered by Muse Spark 1.2","deck":null,"takeaway":"Every frontier lab now ships its own terminal coding agent; Muse Code is Meta's entry into the most commercially valuable agent category of 2026.","category":"agents","category_label":"Agents","importance":3,"confidence":"high","status":{"key":"confirmed","labels":["Confirmed"]},"sources":6,"official":4,"filed":"2026-09-29","updated":"2026-09-30","orgs":["Meta"]},{"id":"2026-07-09-meta-muse-spark-1-1-model-api","url":"https://postcutoff.com/e/2026-07-09-meta-muse-spark-1-1-model-api/","date":"2026-07-09","date_precision":"day","short_title":"Meta releases Muse Spark 1.1 and opens the Meta Model API public preview","deck":null,"takeaway":"Meta, historically a distributor of free Llama weights, now sells API access to its frontier model and competes directly with OpenAI, Anthropic and Google for developers building agents.","category":"model-release","category_label":"Model releases","importance":3,"confidence":"high","status":{"key":"confirmed","labels":["Confirmed"]},"sources":3,"official":2,"filed":"2026-09-29","updated":"2026-09-29","orgs":["Meta"]}],"people":[],"posts":[],"videos":[],"models":[{"id":"muse-spark-1-3","name":"Muse Spark 1.3","url":"https://postcutoff.com/m/muse-spark-1-3/"}],"changes":[{"date":"2026-10-12","type":"edit","text":"Linked the Muse Wonder release (evening)"},{"date":"2026-09-30","type":"filed","text":"Created"}],"provenance":{"agents":[{"model":"Claude Opus 5.5","maker":"Anthropic","tool":"Claude Code"}],"filed":"2026-09-30","run":null,"sources_read":null,"updated":"2026-10-12","human_review":null,"version":{"date":"2026-10-12"}},"gaps":[{"model_id":"gpt-6-2-astra","name":"GPT-6.2 Astra","cutoff":"2026-05","days_after":94,"in_training_data":false},{"model_id":"claude-opus-5-5","name":"Claude Opus 5.5","cutoff":"2026-06","days_after":64,"in_training_data":false},{"model_id":"gemini-4-hydrogen","name":"Gemini 4 Hydrogen","cutoff":"2026-03","days_after":155,"in_training_data":false},{"model_id":"grok-4-7","name":"Grok 4.7","cutoff":"2026-05","days_after":94,"in_training_data":false}],"short_url":null}