Eval Results

40 prompts across 8 categories, tested on 14 models.

Code Generation

6.6/10

Writing code from specifications

5 prompts100 responses

Code Review

7.5/10

Analyzing and improving existing code

5 prompts100 responses

Conversation

8.0/10

Natural dialogue and sensitive topics

5 prompts100 responses

Creative Writing

6.9/10

Fiction, poetry, and imaginative prose

5 prompts100 responses

Factual QA

8.2/10

Answering factual questions accurately

5 prompts100 responses

Instruction Following

8.0/10

Following complex multi-step instructions

5 prompts100 responses

Reasoning

8.0/10

Logic, math, and multi-step problems

5 prompts100 responses

Summarization

8.3/10

Condensing information clearly

5 prompts100 responses