Needle in a Needlestack Code
Llama 3.1 8B: Excels in 8K Contexts, Challenged by Expansion
Jamba 1.5: New model with new architecture crushes Needle-in-a-Needlestack
GPT4o-mini comparable to GPT-4 Turbo, for 98.5% lower price
Sonnet 3.5 Does Much Better at NIAN Than 3.0
Gemini 1.5 Flash Outperforms Much More Expensive Models
GPT-4o’s Memory Breakthrough!