Skip to content
TheChatArchiveThe Chat Archive
gpt-4.1-2025-04-14· OpenAIHugging Face LMArena Human Preference 140K

Whats the best LLM for creative writing, specifically role play scripts

From a Chatbot Arena battleThe same person sent this prompt to two anonymous models at once. Compare it with Coherecommand-a-03-2025’s answer.
  1. User

    Whats the best LLM for creative writing, specifically role play scripts

  2. Assistant

    Great question! The best large language model (LLM) for creative writing and role play scripts depends on your specific needs (e.g., style, interactivity, cost, privacy). As of mid-2024, here’s a breakdown of the top options and their strengths for this use case:


    1. OpenAI GPT-4 (ChatGPT)

    • Strengths:
      • Excellent at creative writing, dialogue, and character development.
      • Can maintain context over long conversations, which is great for role play.
      • Supports custom instructions and memory (with ChatGPT Plus/Team/Enterprise).
    • Weaknesses:
      • May sometimes refuse certain mature or niche role play scenarios due to safety filters.
      • Requires internet access and may have usage limits.

    2. Anthropic Claude 3 (Opus, Sonnet, Haiku)

    • Strengths:
      • Very strong at nuanced, in-character dialogue and maintaining tone.
      • Often more “willing” to engage in imaginative or speculative scenarios.
      • Handles long context windows (up to 200K tokens in Opus).
    • Weaknesses:
      • May be less “edgy” or risk-taking than GPT-4 in some creative scenarios.
      • Access may be limited depending on region/platform.

    3. Google Gemini (Ultra, Pro)

    • Strengths:
      • Good at creative tasks and can generate diverse, interesting scripts.
      • Integrates well with Google Workspace for collaborative writing.
    • Weaknesses:
      • Sometimes less consistent in character voice than GPT-4 or Claude.
      • Context window is smaller than Claude Opus.

    4. Open-Source Models (e.g., Llama-3, Mixtral, Mistral)

    • Strengths:
      • Can be fine-tuned for specific genres, characters, or role play styles.
      • No usage limits or censorship if self-hosted.
      • Great for privacy and custom workflows.
    • Weaknesses:
      • Out-of-the-box creativity and coherence may lag behind GPT-4/Claude.
      • Requires technical setup and possibly fine-tuning for best results.

    Best Overall for Role Play Scripts:

    For most users:

    • GPT-4 (ChatGPT Plus/Team/Enterprise) is the most versatile and creative, with strong dialogue and scene-setting abilities.

    For long, complex role plays:

    • Claude 3 Opus is excellent due to its massive context window and nuanced writing.

    For privacy or niche scenarios:

    • Llama-3 70B (or similar open-source models) with custom fine-tuning is best, especially if you want to avoid content filters.

    Tips for Better Role Play Scripts:

    • Prompt engineering matters: Give clear character backgrounds, tone, and scene details.
    • Iterate: Use follow-up prompts to refine dialogue and actions.
    • Try different models: Sometimes, switching models for different characters/scenes yields richer results.

    Summary Table:

    ModelCreativityDialogueContext LengthCustomizationPrivacy
    GPT-4⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐
    Claude 3 Opus⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐
    Gemini Ultra⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐
    Llama-3 70B⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐⭐

    If you want a sample prompt or script, let me know your genre or scenario!

Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.