There is no publicly available information about OpenAI's GPT-5. However, Anthropic has released Claude 3.5 Sonnet, which they claim outperforms OpenAI's GPT-4o in key evaluations[2]. Anthropic's Claude models utilize Constitutional AI and harmlessness training, which aims to make the AI helpful, honest, and harmless[1].
Claude Models
* Claude 3.5 Sonnet Anthropic says Claude 3.5 Sonnet is their most intelligent model yet and it outperforms competitor models on key evaluations[2]. It is also faster than Claude 3 Opus, but more cost-effective[2][3]. Anthropic is offering free access to Claude 3.5 through the website claude.ai and over the Claude iOS app[2].
* Claude 3 Opus This is a powerful AI model that demonstrates strong performance on highly complex tasks and can navigate open-ended prompts with human-like understanding[7].
* Claude 3 Haiku This is Anthropic's fastest vision and text model, designed for providing near-instant responses to simple queries, and meant to mimic human interactions[7].
Constitutional AI (CAI)
Anthropic uses Constitutional AI (CAI) framework, a training technique developed to align AI systems with human values into large language models (LLMs) like Claude[5]. The core idea behind CAI is to create an explicit "constitution" or set of instructions that encode high-level objectives, ethics, and norms that the AI model should follow[5]. During the supervised learning phase, the model is trained to critique and revise its own outputs based on the principles in the constitution[5].
Citations:
[1] https://help.promptitude.io/en/articles/8892919-understanding-anthropic-models-a-simple-guide
[2] https://www.pcmag.com/news/anthropic-our-claude-35-model-beats-openais-gpt-4o
[3] https://builtin.com/articles/anthropic
[4] https://fortune.com/2024/06/20/anthropic-ai-model-openai-rivalry-continues/
[5] https://www.voiceflow.com/articles/anthropic-ai
[6] https://observer.com/2024/06/anthropic-release-claude-ai-model-gpt-comparison/
[7] https://cloud.google.com/solutions/anthropic
[8] https://www.activeloop.ai/resources/how-to-compare-large-language-models-gpt-4-3-5-vs-anthropic-claude-vs-cohere/