Eval Results

40 prompts across 8 categories, tested on 9 models.

Code Generation

5.4/10

Writing code from specifications

5 prompts45 responses

Code Review

6.3/10

Analyzing and improving existing code

5 prompts45 responses

Conversation

7.4/10

Natural dialogue and sensitive topics

5 prompts45 responses

Creative Writing

6.4/10

Fiction, poetry, and imaginative prose

5 prompts45 responses

Factual QA

7.4/10

Answering factual questions accurately

5 prompts45 responses

Instruction Following

6.6/10

Following complex multi-step instructions

5 prompts45 responses

Reasoning

7.5/10

Logic, math, and multi-step problems

5 prompts45 responses

Summarization

7.5/10

Condensing information clearly

5 prompts45 responses