- Start
- Sep 2, 202690% CONFIDENCEfrom the source
Muse Spark 1.3 Released
- Meta released Muse Spark 1.3 on 2 September 2026, saying it "delivers improved performance across agentic and coding tasks" and draws on "months of broad adoption of Muse Code and Meta Model API"[1]
- It is Meta's fourth Muse Spark release in five months; the variant available now, Muse Spark 1.3 (xhigh), scores 61 on the Artificial Analysis Intelligence Index, up from 57 for Muse Spark 1.2 in August, tying GPT-5.6 Sol (max) and Grok 4.6 (high)[4]
- Muse Spark 1.3 (max) scores 62, behind only Claude Fable 5.1 and Claude Opus 5; Meta's post says max reasoning is available on Muse Code and Meta Model API, while Artificial Analysis described it as a limited preview for Meta's partners and VentureBeat as still in safety testing[1][3][4]
- Meta trained it to ask clarifying questions when a prompt is ambiguous, ask the user for help when stuck and confirm before consequential actions, and to know better "what it can and can't do" instead of hallucinating outcomes[1]
- In comparisons by Meta engineers it used about 20% fewer tool calls and 25% fewer tokens than Muse Spark 1.2[1]
- Meta's roadmap includes "bigger models, the Muse Spark open weights release, and more"[1][3]
Notable features
- Price unchanged from Muse Spark 1.2: $1.25 per million input tokens, $0.15 per million cached input tokens and $4.25 per million output tokens; a contributor tier costs $0.10 and $0.20 per million, with data used to improve Meta's products[2][4]
- A 1M-token context window, with text, image and video input[2][4]
- Artificial Analysis scores: Tau3-Bench Banking 47% for xhigh and 52% for max (the top score of any model) against 35% for Muse Spark 1.2, Terminal-Bench 2.1 85% against 80%, and GDPval-AA v2 1,709 and 1,754 Elo against 1,615[4]
- At $0.55 per Intelligence Index task it is the cheapest model scoring 59 or more; GPT-5.6 Sol (max) costs $0.95 and Grok 4.6 (high) $0.94[4]
- Small regressions: AA-LCR fell from 83% to 79%, and AA-Omniscience accuracy dipped because it declines to answer more often when unsure[4]
- Safety: stronger resistance to adversarial inputs and prompt injections, and better judgement about which agentic actions are irreversible[1]
- Available in Muse Code, Meta's multi-agent coding tool for the macOS and Linux terminal, and the Meta Model API[1]
References 488% CONFIDENCEOverall confidence: 88%How well the pin's source and references back up its dates.Weighted average of how firmly 4 references, the source included, support the pin's start and end times; a reference counts half as much for every 180 days older than the newestShow all pins at 75% confidence or better
The first entry is always the pin's source. Overall confidence is a weighted average of how firmly each reference supports the start and end times used above; a reference counts half as much for every 180 days older than the newest.
- [1]90%research.meta.ai/blog/introducing-muse-spark-1-3research.meta.ai· Posted Sep 28, 2026· Starts Sep 2, 2026 ✓· 26% of score
Meta's[2] post is dated "September 2, 2026" and says "We're excited to release Muse Spark 1.3" and that it "is available today in Muse Code and in Meta Model API"; Artificial Analysis[4]' write-up of the same day opens "Meta has released Muse Spark 1.3".
- [2]90%Muse Spark 1.3 | Metadeveloper.meta.com· Added Sep 28, 2026· 26% of score
Meta's[1] developer model page lists the two API models and their prices - muse-spark-1.3 at $1.25 input, $0.15 cached input and $4.25 output per million tokens, and muse-spark-1.3-contributor at $0.10 and $0.20 with data used to improve Meta's products - both with a 1M-token context window.
- [3]80%Meta says Muse Spark 1.3 has frontier performance - but its best results come from a model developers can't broadly use yetventurebeat.com· Published Sep 3, 2026· 24% of score
VentureBeat's report by Carl Franzen says the xhigh version is broadly available through Muse Code and Meta[1][2] Model API while the max configuration is undergoing safety testing and will arrive shortly, and notes the open-weights release now on Meta's roadmap.
- [4]90%Muse Spark 1.3: Meta reaches the frontierartificialanalysis.ai· Published Sep 2, 2026· Starts Sep 2, 2026· 24% of score
Artificial Analysis' same-day independent benchmark write-up: Muse Spark 1.3 (xhigh) scores 61 on its Intelligence Index, up from 57 for 1.2 and 53 for 1.1, and the limited-preview max variant 62, behind only Claude Fable 5.1 and Claude Opus 5; it gives the Tau3-Bench Banking, Terminal-Bench 2.1, GDPval-AA v2 and cost-per-task figures.
Suggest a correction
Something missing or wrong? Say it in your own words: a link that backs this pin up, a different start or end date and why, or a fact it lacks or gets wrong. The AI checks it against this pin's sources, searches for better ones, and adds any page that backs you up. The pin's own sources still count most. A picture that shows something else, or shows it badly, is looked at too, and moved down or replaced.
