Taalas just introduced the HC1, an AI accelerator that bakes the Llama 3.1 8B model directly into hardware to generate a blistering 17,000 tokens per second. By merging compute and storage on a single chip, it claims 40 times the speed of Nvidia's B200 at a fraction of the cost, though it is permanently tied to that specific model.
OpenAI and Meta are partnering with telecom giant Orange to fine-tune AI models for regional African languages currently ignored by mainstream GenAI. The project starts with West Africa's Wolof and Pulaar, training Whisper and Llama 3.1 to power voice support, with plans to release models under a free non-commercial license for local health and education. Why it matters: It’s a practical push for digital inclusion, bringing AI to millions who don't speak dominant global languages.
Meta just revealed the winners of its inaugural Llama Impact Innovation Awards, awarding organizations across Africa, the Middle East, Asia Pacific, and Latin America up to $35,000 in funding and matching ad credits. The recognized projects leverage Llama models for local challenges, including bilingual medical diagnosis, vaccine education, and low-cost vision aids.
Meta just dropped new adoption stats for Llama, revealing total downloads on Hugging Face are approaching 350 million—a 10x jump compared to last year. Token volume across major cloud partners more than doubled between May and July 2024, with the massive 405B model driving the highest unique users in August. Open-source AI is no longer just for tinkerers; it's rapidly anchoring enterprise tech stacks.
Cerebras just launched its new AI inference service powered by the CS-3 system, clocking an eye-watering 1,800 tokens per second on Llama 3.1 8B. That is 20 times faster than hyperscale Nvidia GPUs while retaining full 16-bit precision at just 10 cents per million tokens. Why it matters: lightning-fast generation is critical for multi-step AI agents to finally feel real-time rather than sluggish.
Meta just rolled out AI Studio in the US, allowing anyone to build and discover custom AI characters powered by Llama 3.1. Creators can set up AI extensions of themselves to answer common questions and engage with followers. However, Meta later noted it has halted new character creation as it shifts focus to Muse Spark.
Meta just dropped its open-source Llama 3.1 family, headlined by a massive 405B model that rivals GPT-4o and Claude 3.5 Sonnet across key benchmarks. The entire lineup supports a 128K context window, and Meta has updated its license so developers can finally use 405B outputs to distill smaller models.
