Meta Launches Muse Spark 1.3, Claims It Now Rivals OpenAI and Anthropic

By Carlos Montiel | Enterprise AI Specialist
Leer en español →
Published: 2026-09-08 | By: Carlos Montiel | Reading time: ~4 min

On September 2, Meta released Muse Spark 1.3, its most powerful model to date, and Chief AI Officer Alexandr Wang described it as "the company's biggest performance leap," positioning it alongside OpenAI's and Anthropic's most recent models.

A jump in coding and agentic tasks, at no extra cost

According to Artificial Analysis, Muse Spark 1.3 scored 62 on its Intelligence Index, trailing only Claude Fable 5.1 and Claude Opus 5, and ahead of the OpenAI models evaluated. Wang said the model is "competitive" with Claude Fable 5.1 and "better" than OpenAI's GPT-5.6 Sol, particularly on coding tasks. On efficiency, Meta reports that Muse Spark 1.3 uses roughly 25% fewer tokens than version 1.2 to complete equivalent tasks, with roughly 20% fewer tool calls according to engineering tests the company reported.

Meta Muse Spark 1.3 — verified data Release date: September 2, 2026 Intelligence Index (Artificial Analysis): 62 — 3rd place, behind Claude Fable 5.1 and Claude Opus 5 Efficiency vs. Spark 1.2: ~25% fewer tokens, ~20% fewer tool calls Price: unchanged from Spark 1.2 Context window: 1M tokens Availability: Meta Model API (developers) + Facebook, Instagram and Meta AI app (consumer)

Price frozen, open weights still undecided

Meta kept Muse Spark 1.3's price the same as Spark 1.2's, positioning it as "one of the most accessible LLMs on the market." API access is already available for developers paying by usage, while rollout to Facebook, Instagram and the Meta AI app users will arrive in the coming days. Meta hasn't yet decided whether it will release the model's weights openly, a departure from its historical strategy with the Llama family.

The absence of a price increase despite the performance gain is the most relevant signal for AI procurement teams: Meta is competing aggressively on price-performance against OpenAI and Anthropic, which pushes down cost-per-token across the entire frontier-model market.

What it means for a company choosing a model

Meta landing, by its own metrics and Artificial Analysis's, in third place on the intelligence ranking — behind Claude Fable 5.1 and Claude Opus 5, and ahead of the OpenAI models evaluated — confirms that the competitive gap between frontier labs keeps narrowing every few weeks. For a company that has already built its stack around a specific model, this isn't a reason to migrate immediately, but it is a reason to keep the LLM architecture decoupled from the provider, so that switching models is a configuration decision rather than a rewrite.

For companies in Guatemala and Latin America evaluating AI providers for coding tasks or high-volume agentic workflows, Muse Spark 1.3 is worth piloting: same price as its predecessor, fewer tokens consumed per equivalent task, and immediate availability via API — no waiting on restricted-access approvals like some other recent frontier models require.
Carlos Montiel
Enterprise AI Solutions Architect
LLMs, Agents & Orchestration Specialist
guatemalia.com/#contacto · info@guatemalia.com

Need to implement AI in your company?

Carlos Montiel is an enterprise AI solutions architect. He implements LLMs, Agents, RAG and orchestrators for companies across Guatemala and Latin America. Reach out for a consultation.

Contact Carlos Montiel

info@guatemalia.com