Meta debuts new AI model, attempting to catch Google, OpenAI after spending billions
google meta openai
| Source: CNBC | Original article
Meta Platforms unveiled its first flagship large‑language model, Muse Spark, on Wednesday, positioning the company for a direct showdown with Google’s Gemini and OpenAI’s GPT‑4. The announcement, made by chief AI officer Alexandr Wang, highlighted Muse Spark’s multimodal capabilities—text, image and audio processing—built on Meta’s proprietary LLaMA‑2 architecture and optimized for the company’s massive data pipelines. The model will initially power a new “Meta AI” chatbot and a suite of over two dozen AI‑generated characters slated for Facebook, Instagram and WhatsApp.
The debut matters because it marks Meta’s transition from incremental AI features to a core, proprietary foundation model that can be monetized across its ad‑driven ecosystem. By keeping the model in‑house, Meta aims to reduce reliance on external providers, cut licensing costs and tighten control over data privacy—a recurring concern for European regulators. Muse Spark also signals that Meta is finally capitalising on the billions it has spent on AI research, a spend that has lagged behind rivals in public perception.
The rollout arrives after Meta postponed its previously teased “Avocado” model, which was pushed back from a March launch to at least May following internal testing setbacks, according to the New York Times. Muse Spark therefore serves as a stop‑gap to demonstrate progress while the Avocado model is refined.
What to watch next: independent benchmarks will reveal whether Muse Spark can match or exceed the accuracy and latency of Gemini and GPT‑4. Developers will be keen for Meta’s API access timeline, and advertisers will monitor how the model integrates with targeting tools. Finally, the performance of the delayed Avocado model will indicate whether Meta can sustain a rapid innovation cadence in the fiercely competitive generative‑AI race.
Sources
Back to AIPULSEN