Meta released Llama 3.1 405B in July 2024, positioning it as the first frontier-level open-source AI model competitive with closed proprietary offerings. Zuckerberg claims inference cost approximately 50 pct below GPT-4o for both user-facing and offline tasks, and notes that the open architecture enables enterprise fine-tuning and distillation without model provider data access. Meta framed the release as part of long-term ecosystem strategy, partnering with AWS, Azure, Google, Oracle, Databricks, NVIDIA, and others to standardize on open-weight infrastructure. For financial institutions, this signals a structural shift in cost-of-inference leverage and enterprise optionality: enterprises can now run frontier-parity inference at lower marginal cost and without exclusive vendor lock-in, which pressures proprietary API pricing and shifts economic rent toward infrastructure and systems integration.
244 other tracked developments touching the same claims.