PitchBook: 95+ investors put money into both Anthropic and OpenAI amid shifting norms; sources: Menlo, Lightspeed, and Iconiq each hold 1%-2% Anthropic stakes (Erin Griffith/New York Times)

Meta has unveiled its latest AI model, Muse Spark 1.3, boasting a significant leap in performance compared to its predecessors and the leading models from OpenAI and Anthropic. Developers can now access and pay for this cutting-edge update, which will soon be integrated into Meta's popular apps such as Instagram, Facebook, and Meta AI.

Meta's Chief AI Officer Alexandr Wang touted this update as the company's "biggest jump so far on model performance," highlighting substantial improvements in coding and agentic tasks – the AI's capacity to perform tasks on behalf of users. Wang asserted that Muse Spark 1.3 stands toe-to-toe with Anthropic's Claude Fable 5.1 and even outperforms OpenAI's GPT-5.6 Sol in coding capabilities, though OpenAI's upcoming Astra model remains in the pipeline.

While these benchmark comparisons should be taken with a pinch of caution, as models can be optimized for specific tasks, Wang elaborated on several upgrades that distinguish Muse Spark 1.3. This new model requires 25% fewer tokens to complete tasks, resulting in enhanced efficiency. Additionally, it can manage multiple workflows simultaneously, eliminating the need for separate sessions.

Moreover, Muse Spark 1.3 demonstrates a better understanding of its own limitations, pausing to request clarification before undertaking actions that cannot be reversed.

At this time, it remains unclear whether Meta will release the model's weights, the building blocks that enable other developers to construct upon its foundation. However, the older Muse Spark 1.2 weights are slated for release shortly. Pricing remains unchanged for developers, as they will not face increased costs for Muse Spark 1.3 compared to its predecessor.

Despite Meta's ongoing investment in AI development, Wang emphasized that adoption on Meta's developer platform has been robust, with some users consuming trillions of tokens weekly.

Meta's AI ambitions continue to expand, but with the introduction of Muse Spark 1.3, the company is sending a clear message that it is closing the gap with its competitors. Anthropic recently made waves with an upgrade to Claude Cowork and Claude Code, enabling the AI to operate discreetly in the background on Mac devices. This feature allows Claude to perform tasks without obstructing the user's screen.

However, researchers from Product Rise discovered that when the same product appears in both Claude Cowork and traditional Google Search, the AI-powered experience tends to promote products at a 21.6% higher average price. The study compared Claude's pricing against traditional Google Search, finding that AI Mode often exhibited inflated prices for products.

Moreover, Amazon has become increasingly vigilant in combating fraudulent shopping-related emails and messages, which have become increasingly sophisticated. The company now fields an average of 360,000 inquiries annually, attempting to discern genuine correspondence from deceptive messages. Starting today, customers can rely on Alexa to verify the authenticity of suspicious shopping-related emails and messages, simplifying the process of confirming the legitimacy of such communications.