I’ve even found this to be true on reasoning tasks, which is shocking
@mattshumer_
-
Few-Shot Prompting Technique Boosts Claude 3 Haiku Performance
By
–
Highest alpha secret in AI right now: If you provide ~10 examples to Claude 3 Haiku… it’ll often outperform Claude 3 Opus, and far outperform GPT-4 at a fraction of the cost, with blazing fast speeds
-
Claude 3 Opus Outperforms GPT-4 on Debugging Task
By
–
Gotta hand it to @AnthropicAI
, Claude 3 Opus is an absolute beast of a model I've was trying to get GPT-4 to debug an issue for two hours, no success Same prompt, Opus gets it on the first try -
HyperWrite AI Writing Tool Launches with Custom Model
By
–
We have very high limits once you subscribe (depends on the plan but more than enough for most users). Base plan is $20/mo USD. HyperWrite writes far more naturally than GPT-4, as we use our own in-house model that is trained primarily for writing. Once you try it, you'll see!
-
Integrating Metaprompt Framework for Enhanced AI Results
By
–
This should provide better results. But it’d be very easy to integrate Metaprompt into this framework for even better results!
-
AI System Shows Notable Improvements in Performance Capabilities
By
–
It’s noticeably better, smarter, more thorough and malleable
-
Direct API usage over LLM frameworks for faster development
By
–
The reason I can work so fast is simple: I use no LLM frameworks, just hit the APIs directly for what I need. Frameworks are a time-suck. That, and I use Claude as an engineering assistant. While I’m working on writing a prompt or integrating some code, I’ll simultaneously have
-
Claude 3 Instance Evaluates Multiple Outputs Decision Making
By
–
A separate Claude 3 prompt instance is shown the two outputs, the task description, inputs, etc., and can make a decision from there. Lots of ways to make this more robust.