First impressions of Amazon Nova LLMs: they are competitive with Google Gemini, among the cheapest available, and may position Amazon as a top model provider (@simonw / Simon Willison's Weblog) https://t.co/vhUhg1ttVo https://t.co/Kg7i7QfXVB https://t.co/ZOzeer2dpR
Could we please talk about Amazons new Nova models for a second? Let's start with Nova Micro: 128k context length, 5k output, text only model currently priced at $0.04 Input, $0.14 Output (per 1M tokens) It's direct competitor is Gemini 1.5 Flash 8B. Which currently supports… https://t.co/MKtMDlamHb
I spent some time yesterday exploring the new Amazon Nova LLM family, and I'm really impressed They're price and quality competitive with Google Gemini - and Nova Micro is now the cheapest model from any of the major vendors (cheaper even than Gemini 1.5 Flash-8B) https://t.co/Kq6hGmrWp0
Amazon has launched a new family of models called Nova, which have received mixed reviews based on their performance in the LiveBench AI benchmarks. In the latest November challenge, Nova scored below Llama-70b but slightly outperformed Haiku. The Nova Pro model has achieved a fourth-place ranking on the updated LiveBench leaderboard, with Claude 3.6 Sonnet still holding the top position. The Nova models are noted for being competitively priced, particularly the Nova Micro, which offers a context length of 128k and an output of 5k, priced at $0.04 per million tokens for input and $0.14 for output. This pricing makes it the cheapest model among major vendors, even undercutting Google’s Gemini 1.5 Flash-8B. Observers have highlighted that Nova's introduction marks a significant improvement for Amazon, especially in light of previous setbacks with its AI models, which resulted in a failed attempt using a 14,000 A100 cluster nine months ago.