meta/muse-spark-1.3

模型发布 OpenRouter 模型 原文语言:中文

OpenRouter 于 2026-09-03 收录该模型(收录时间不等于模型正式发布日期);上下文:1,048,576 tokens;输入价:$1.2500/百万 tokens;输出价:$4.2500/百万 tokens;定位(来源描述):Muse Spark 1.3 is a multimodal reasoning model from Meta for long-running agentic, multi-agent,…

全文

Meta: Muse Spark 1.3

meta/muse-spark-1.3

Compare

Muse Spark 1.3 is a multimodal reasoning model from Meta for long-running agentic, multi-agent, and coding workflows. It is designed to keep track of information across extended tasks, work through conflicting inputs, and request clarification or confirmation when needed, with an emphasis on concise execution.

Modalities

In / out price

$1.25 / $4.25per 1M

Context

1.0M

Released

Sep 2, 2026

Compare

Providers

Different companies host the same model. OpenRouter routes your request to one of them based on the routing mode you pick: Balanced (price + speed), Nitro (fastest), or Floor (cheapest).

Latency / throughput

$1.25$4.25$0.151.97s166 tok/s

99.91%

$1.25$4.25$0.153.73s101 tok/s

99.97%

ProviderInput / Output /M

Input $1.25 / Output $4.25

166 tok/s99.91% uptime

Input $1.25 / Output $4.25

101 tok/s99.97% uptime

Pricing

The average price customers actually pay for this model, next to the prices providers post. Caching and discounts mean the price actually paid is often well below the listed one.

Weighted average

Input

$0.3674/M tokens

Output

$4.25/M tokens

Muse Spark 1.3 — Price history

Price history

No effective price data for this selection

$0.331$4.2583.5%64.8%

$0.434$4.2574.2%35.2%

ProviderInput / Output /M

Token share

Effective in $0.331 / Effective out $4.25

Token share

64.8%

Effective in $0.434 / Effective out $4.25

Token share

35.2%

Performance

Throughput is how fast the model writes (tokens per second — higher is better). Latency is total round-trip time (lower is better). TTFT is time-to-first-token — how long before you see anything appear (lower is better).

Throughput

166tok/s

P50, best across providers

Latency

1.97s

P50, best provider

Reasoning effortLatency / throughput

Uptime

Uptime is the percentage of the past 3 days that at least one provider was responding to requests. Availability is the percentage of time that inference was successfully served. OpenRouter continuously monitors and uses the next-best provider when one returns an error.

Uptime (3d)The model was reachable. Request routed to a provider.

100.00%

Availability (3d)The model returned inference from any provider. Errors and empty responses count against it.

99.71%

Availability over the last 3 days

Last 72 hours

Availability 99.71%

3 Days Ago2 Days AgoYesterdayNow

Availability over the last 24 hours

OpenRouter Availability

99.73%

Without Routing

98.86%

When an error occurs in an upstream provider, we can recover by routing to another healthy provider, if your request filters allow it. You can access per-provider uptime data programmatically through the Endpoints API. Learn more about our load balancing and provider routing options.

Benchmarks

Scores on standardized evaluations and this model’s own outputs on shared benchmark prompts. Higher percentages are better — and each rank shows where this model lands among the models evaluated on that source.

Benchmark score summary for Meta: Muse Spark 1.3 (Artificial Analysis, Design Arena and Vals AI)

SourceBenchmarkScore

Artificial AnalysisMuse Spark 1.3 (Xhigh) Coding Index76.5

Artificial AnalysisMuse Spark 1.3 (Xhigh) Agentic Index51.5

Artificial AnalysisMuse Spark 1.3 (Xhigh) GPQA Diamond94.1%

Artificial AnalysisMuse Spark 1.3 (Xhigh) HLE47.5%

Artificial AnalysisMuse Spark 1.3 (Xhigh) AA-LCR83.0%

Artificial AnalysisMuse Spark 1.3 (Xhigh) τ-Bench Banking47.2%

Artificial AnalysisMuse Spark 1.3 (Xhigh) GDPval-AA56.5%

Artificial AnalysisMuse Spark 1.3 (Xhigh) CritPt26.0%

Artificial AnalysisMuse Spark 1.3 (Xhigh) SciCode59.7%

Artificial AnalysisMuse Spark 1.3 (Xhigh) Terminal-Bench 4.016.7%

Artificial AnalysisMuse Spark 1.3 (Xhigh) Terminal-Bench 2.185.4%

Artificial AnalysisMuse Spark 1.3 (Xhigh) AA-Omniscience Accuracy37.8%

Artificial AnalysisMuse Spark 1.3 (Xhigh) AA-Omniscience Non-Hallucination Rate77.5%

Artificial AnalysisMuse Spark 1.3 (Max) Intelligence Index36.6

Artificial AnalysisMuse Spark 1.3 (Max) Coding Index75.8

Artificial AnalysisMuse Spark 1.3 (Max) Agentic Index55.5

Artificial AnalysisMuse Spark 1.3 (Max) GPQA Diamond93.5%

Artificial AnalysisMuse Spark 1.3 (Max) HLE48.7%

Artificial AnalysisMuse Spark 1.3 (Max) AA-LCR83.0%

Artificial AnalysisMuse Spark 1.3 (Max) τ-Bench Banking50.5%

Artificial AnalysisMuse Spark 1.3 (Max) GDPval-AA59.0%

Artificial AnalysisMuse Spark 1.3 (Max) CritPt24.9%

Artificial AnalysisMuse Spark 1.3 (Max) SciCode58.8%

Artificial AnalysisMuse Spark 1.3 (Max) Terminal-Bench 4.033.3%

Artificial AnalysisMuse Spark 1.3 (Max) Terminal-Bench 2.184.3%

Artificial AnalysisMuse Spark 1.3 (Max) AA-Omniscience Accuracy38.0%

Artificial AnalysisMuse Spark 1.3 (Max) AA-Omniscience Non-Hallucination Rate74.8%

Design ArenaMuse Spark 1.3 Max Agents Arena Androidnative Elo1301

Design ArenaMuse Spark 1.3 Max Agents Arena Blender3d Elo1145

Design ArenaMuse Spark 1.3 Max Agents Arena Codetovideo Elo1207

Design ArenaMuse Spark 1.3 Max Agents Arena Full Stack Elo1287

Design ArenaMuse Spark 1.3 Max Agents Arena Godotgamedev Elo1216

Design ArenaMuse Spark 1.3 Max Agents Arena Mobile Apps Elo1210

Design ArenaMuse Spark 1.3 Max Agents Arena Pdf Elo1184

Design ArenaMuse Spark 1.3 Max Agents Arena Spreadsheet Elo1118

Design ArenaMuse Spark 1.3 Max Agents Arena Webapps Elo1285

Design ArenaMuse Spark 1.3 Max Models Arena 3D Elo1380

Design ArenaMuse Spark 1.3 Max Models Arena Asciiart Elo1330

Design ArenaMuse Spark 1.3 Max Models Arena Code Categories Elo1353

Design ArenaMuse Spark 1.3 Max Models Arena Data Visualization Elo1306

Design ArenaMuse Spark 1.3 Max Models Arena Game Development Elo1345

Design ArenaMuse Spark 1.3 Max Models Arena SVG Elo1314

Design ArenaMuse Spark 1.3 Max Models Arena UI Component Elo1352

Design ArenaMuse Spark 1.3 Max Models Arena Website Elo1353

Vals AIMuse Spark 1.3 Access to Justice Benchmark72.2%

Vals AIMuse Spark 1.3 Code Migration27.6%

Vals AIMuse Spark 1.3 CyberBench v1.169.4%

Vals AIMuse Spark 1.3 EMB62.7%

Vals AIMuse Spark 1.3 Finance Agent (v2)58.9%

Vals AIMuse Spark 1.3 Financial Coaching Bench71.0%

Vals AIMuse Spark 1.3 Harvey's Legal Agent Benchmark v122.9%

Vals AIMuse Spark 1.3 Immigration65.0%

Vals AIMuse Spark 1.3 IOI43.9%

Vals AIMuse Spark 1.3 Legal Research Bench40.9%

Vals AIMuse Spark 1.3 ProgramBench0.5%

Vals AIMuse Spark 1.3 ProofBench v1.155.0%

Vals AIMuse Spark 1.3 Tax Agent Bench37.9%

Vals AIMuse Spark 1.3 Terminal-Bench 4.010.6%

Vals AIMuse Spark 1.3 Terminal-Bench-Science4.3%

Vals AIMuse Spark 1.3 Vals Index53.2%

Vals AIMuse Spark 1.3 Vibe Code Bench v1.182.9%

Apps

Public apps that send the most traffic to this model. Good signal for what real production workloads look like — and a hint at which use cases this model is best suited for.

1.

Hermes Agent

Hermes Agent is an open-source, self-improving AI agent by Nous Research that runs persistently with memory across sessions, and builds reusable skills from experience. It comes with 40+ built-in tools, including web search, browser automation, and vision, plus scheduled automations and subagents.

24.9Btokens

2.

Chat AI App (image-rater)

new

15Btokens

3.

omp

new

9.81Btokens

4.

Claude Code

Claude Code is Anthropic's agentic coding tool that reads your entire codebase, plans and executes changes across files, runs tests, and iterates on failures, all from natural language prompts.

7.97Btokens

5.

Codex

A coding agent that helps you build and ship with AI

5.45Btokens

Rest of OpenRouter

2.6B

Top 5 Apps

471M

Activity

Token volume and request traffic to this model over time.

API

Drop-in code to call this model. OpenRouter's API is OpenAI-compatible — most SDKs work by just swapping the base URL. The only thing that changes between models is the model slug below.

Not currently available through EU in-region routing. In-region routing docs

Frequently asked questions

Muse Spark 1.3 is a multimodal reasoning model from Meta for long-running agentic, multi-agent, and coding workflows. It is designed to keep track of information across extended tasks, work through conflicting inputs, and request clarification or confirmation when needed, with an emphasis on concise execution.

Muse Spark 1.3 costs $1.25/M input tokens and $4.25/M output tokens, with separate rates for Cache Read at $0.15/M tokens.

Muse Spark 1.3 has a 1,048,576 token context window.

Yes. Muse Spark 1.3 accepts tools and tool_choice for function calling. It also supports structured outputs via a JSON schema in response_format.

Muse Spark 1.3 accepts text, images, video, and files such as PDFs as input and returns text.

Muse Spark 1.3 is served by 2 providers on OpenRouter: Google Vertex and Meta. Requests are routed to the best available provider, with automatic failover to the others, and you can pin or exclude providers with provider routing.

Muse Spark 1.3 was released on September 2, 2026.

正文以来源语言显示。

参考来源

← 返回每日简报