Phind AI Model
Phind AI Latest Model
Phind-70B
Phind-70B is built on CodeLlama-70B and is good capacity having been tuned with ~50 billion extra tokens. With a 32K token context window, it can solve complicated long code problems, and being fast on deep technical coding, it produces 80 tokens a second.
Phind-405B
Phind-405B is a flagship large model, built on Llama 3.1 405B. It’s designed for maximum context understanding at launch, it supports a 32K token window, with training for up to 128K tokens.
Phind Instant
Phind Instant is the lightweight fast model designed for rapid query responses. Based on Llama 3.1 8B, it runs at up to 350 tokens per second.
Phind-CodeLlama-34B-v2
Phind-CodeLlama-34B-v2 model is tuned from CodeLlama-34B and optimized for coding tasks. It has a pass@1 score of 73.8% on HumanEval, supports languages like Python, Java, C/C++.