Meta AI Expands With Muse Spark 1.3 and Muse Voice Transcribe

Meta is expanding its artificial intelligence portfolio with two new products aimed at developers and users who need more advanced AI capabilities. The company has introduced Muse Spark 1.3, a model focused on coding and agentic tasks, alongside Muse Voice Transcribe, a real-time speech recognition model.

The launches highlight Meta’s continued push to compete in the rapidly developing AI market, where companies are increasingly focusing on coding, reasoning, voice technology and AI agents.

Muse Spark 1.3 Focuses on Coding and AI Agents

The first major release is Muse Spark 1.3, the latest model in Meta’s Muse Spark family.

Meta describes the model as being designed for challenging reasoning and agentic tasks, with improvements aimed at coding and long-horizon workflows. It is available through Muse Code and the Meta Model API, giving developers access to its capabilities for software development and AI applications.

Unlike conventional AI systems that primarily respond to individual prompts, agentic models are designed to handle multiple steps while working toward a particular objective.

Meta AI chief Alexandr Wang has also connected the company’s model development to its broader plans for personal AI agents. Axios reported that Meta sees these systems as potentially capable of working continuously and completing tasks on behalf of users.

Muse Voice Transcribe Brings Real-Time Speech Recognition

Meta’s second major release is Muse Voice Transcribe, a real-time audio perception model developed by Meta Superintelligence Labs.

The system combines streaming automatic speech recognition, speaker diarization and endpointing in a single model. It can distinguish between more than 20 speakers and is designed to process long audio recordings.

Meta says Muse Voice Transcribe supports more than 70 languages, with 25 languages extensively verified at launch. It also supports multilingual conversations and code-switching, allowing users to move between languages during a conversation.

The model is available through the Meta Model API and has also been integrated into Meta AI for Mac and Muse Code.

Meta lists pricing at approximately $0.18 per hour, or $3 per 1,000 audio minutes, making cost another part of the company’s effort to attract developers.

Support for Indian Languages

The launch could also be significant for users in multilingual markets such as India.

Muse Voice Transcribe supports five Indian languages: Hindi, Tamil, Telugu, Malayalam and Kannada. Its real-time capabilities could make the technology useful for applications involving meetings, customer support, accessibility and voice-based software.

Meta says the model ranked first on Artificial Analysis’ streaming speech-to-text leaderboard as of September 1, 2026. Because this is a company-reported benchmark result, it is more appropriate to attribute the ranking to Meta rather than treat it as an independently established industry-wide conclusion.

Meta AI Targets Developers With Competitive Pricing

Pricing has become an important factor in the AI model market as companies compete for developers.

Muse Spark 1.3 reportedly maintains the pricing approach of its predecessor, with Wang describing Meta’s pricing strategy as aggressive. The model is available through Muse Code and the Meta Model API.

Meta is also offering a contributor tier for its coding products, which reduces costs in exchange for allowing Meta to use developers’ work to improve its models. Wang said a meaningful double-digit percentage of coders choose this option.

For developers, lower AI costs can make it easier to integrate models into coding assistants, automated workflows and other applications requiring frequent model usage.

Meta AI Competition Is Intensifying

The latest releases arrive as the AI industry becomes increasingly competitive.

Google, OpenAI and Anthropic are also developing advanced models aimed at reasoning, coding, voice processing and AI agents. This means Meta is competing not only on model performance but also on pricing, developer access and the ability to turn AI research into practical products.

Muse Spark 1.3 and Muse Voice Transcribe show how Meta is expanding its AI portfolio across different use cases rather than relying on a single model.

Legal Settlement Provides Business Context

Meta’s new AI releases also come shortly after the company agreed to a settlement worth up to $18 billion with nearly all U.S. states over claims concerning children’s use of its social media platforms.

Reuters reported that the agreement includes $12.7 billion in guaranteed payments, with up to another $5 billion potentially payable depending on actions by other social media companies.

Some analysts have argued that resolving the litigation could remove a significant legal obstacle and allow Meta to focus more heavily on its product pipeline. However, there is no evidence that the settlement directly caused the launches of Muse Spark 1.3 or Muse Voice Transcribe.

The settlement is therefore better understood as business context surrounding Meta’s broader AI expansion.

Meta’s Push Toward Personal AI

The launch of Muse Spark 1.3 also fits into Meta’s longer-term ambition to develop more capable personal AI systems.

The company’s vision extends beyond chatbots that simply answer questions. Meta is increasingly interested in AI agents that can reason through complex tasks, use tools and potentially operate with greater independence.

Muse Spark’s focus on coding and long-running agentic workflows could provide part of the technical foundation for this direction.

Key Details

Product	Main Focus	Availability
Muse Spark 1.3	Coding, reasoning and agentic tasks	Muse Code and Meta Model API
Muse Voice Transcribe	Real-time speech recognition	Meta Model API, Meta AI for Mac and Muse Code
Language support	70+ languages	Supported integrations
Speaker diarization	20+ speakers	Muse Voice Transcribe
Voice pricing	About $0.18 per hour	Meta Model API

Frequently Asked Questions

What is Muse Spark 1.3?

Muse Spark 1.3 is Meta’s latest AI model focused on coding, reasoning and long-horizon agentic workflows.

What is Muse Voice Transcribe?

Muse Voice Transcribe is a real-time speech recognition model that combines transcription, speaker identification and endpointing.

How many languages does Muse Voice Transcribe support?

Meta says it supports more than 70 languages, with 25 extensively verified at launch.

Is Muse Voice Transcribe API-only?

No. It is available through the Meta Model API and is also integrated into Meta AI for Mac and Muse Code.

What is Meta’s goal with AI agents?

Meta is working toward more capable personal AI systems that can understand goals, reason through tasks and potentially take actions on behalf of users.

Conclusion

The latest Meta AI releases demonstrate the company’s continued effort to compete in an increasingly crowded artificial intelligence market.

Muse Spark 1.3 targets coding, reasoning and agentic workflows, while Muse Voice Transcribe brings real-time multilingual speech recognition to Meta’s developer ecosystem and selected products.

The launches also reflect a wider shift in AI development toward practical capabilities such as coding, voice processing and autonomous task completion. While Meta’s recent legal settlement provides important business context, it should not be treated as the direct cause of these product releases.

For Meta, the bigger challenge will be turning these new technologies into reliable products that developers and users continue to find valuable.