Choosing a Model

Overview

Genesis Writer gives you access to 20+ AI models from every major provider: Anthropic (Claude), OpenAI (GPT), Google (Gemini), xAI (Grok), MoonshotAI (Kimi), and DeepSeek. Each model has its own personality — different strengths, different weaknesses, and different costs.

The Free plan starts you on GPT-5.6 Luna. Any paid subscription unlocks all of them. You can switch models between generations, so there's no commitment — use a cheap model for rough drafts and a premium one for your most important scenes.

The model selector — switch models anytime from the toolbar

Your Free Model Free

The Free plan starts you on GPT-5.6 Luna (OpenAI) — a fast, capable model with a novel-length context window (1M tokens). It's economical, instruction-focused, and consistent across long manuscripts, which makes it well suited to outlining, drafting, and revision alike.

Any paid plan unlocks 20+ models from every major provider — the budget, balanced, premium, and ultra tiers below.

Budget Models Writer+

Affordable subscriber options for everyday writing — fast, capable models at a price that won't drain your token balance.

Claude 4.5 Haiku is the standout here. It has a natural ear for dialogue — the way characters talk feels less like AI than most models at this price point. If your scene is heavy on conversation, Haiku is an excellent choice.

Balanced Models Writer+

The sweet spot between cost and quality. These models produce noticeably better prose than budget options while remaining affordable for regular use.

Premium Models Writer+

The heavy hitters. These models produce the highest-quality prose, handle complex narratives, and offer the most nuanced writing. Use them for scenes that matter.

The two Claude Sonnets have different strengths. 4.5 Sonnet is the emotional writer — it gets inside a character's head convincingly and excels at scenes that need to land emotionally. 4.6 Sonnet is tighter and more controlled — better at holding a specific voice across twenty chapters without drift.

Ultra Models Writer+

The best of the best. Two models, no compromises.

Claude Opus 5 is the closest thing to a human co-writer in the lineup. It understands character psychology — motivations, contradictions, the gap between what characters say and what they mean. Prose has genuine texture and weight.

It's the most expensive model at 5 cost per generation, and it's slower than the rest. Save it for the scenes that matter most — the climactic confrontation, the quiet moment of grief, the chapter that makes or breaks your novel.

Full Comparison Table

Here's every model at a glance. Ratings are on a 1–5 scale. Cost is measured in cost (relative units — see The Token Economy for how cost translate to Genesis Tokens).

Recommendations by Use Case

Not sure which model to pick? Here are battle-tested recommendations for common scenarios:

Fast Drafting

When you want to get ideas down quickly and worry about polish later, reach for Gemini 3.5 Flash Lite or GPT-5 Mini. Both are fast, affordable, and produce coherent prose. GPT-5 Mini can generate very long passages without losing the thread, making it especially good for marathon drafting sessions.

Character Voice & Dialogue

Anthropic's Claude models have the best ear for how people actually talk. Claude 4.5 Haiku is the budget-friendly option — fast and natural-sounding. For deeper character work, Claude 4.5 Sonnet and Claude 4.6 Sonnet produce dialogue that feels genuinely human.

Emotional Scenes

Scenes that need to land emotionally — grief, longing, quiet tension, a character's internal reckoning — are where Claude 4.5 Sonnet and Claude Opus 5 truly shine. They show rather than tell, and they render internal states with surprising depth.

Long Chapters

If you need the AI to generate long, sustained passages (full chapters, extended scenes), choose a model with a high max output: GPT-5 Mini (128K), GPT-5.2 (128K), GPT-5.4 (128K), or GPT-5.6 Luna, Terra, or Sol (128K), or Claude Opus 5 (128K). These models maintain coherence over very long stretches.

Mature & Unrestricted Content

Most models have content restrictions that limit how they handle explicit romance, graphic violence, or taboo themes. If you need a model that writes any scene without hesitation, use Grok 4.3 (premium, Freedom 5/5) — xAI's current flagship. Grok handles mature content natively while still delivering premium-quality reasoning across long arcs.

Literary Fiction

For prose that has genuine texture — subtext, layered meaning, psychological depth — use Claude Opus 5 or Claude Sonnet 5. Opus is the best in the lineup at writing prose that reads like it was crafted by a human author.

Plot-Heavy Fiction

Stories with complex magic systems, interconnected mysteries, or elaborate world rules need a model that thinks structurally. Gemini 3.1 Pro and GPT-5.6 Terra are both excellent at maintaining internal consistency and tracking plot threads.

Budget-Friendly Writing

If you want to maximize your token balance, GPT-5.6 Luna is the cheapest way to write all day — and it's the model every plan includes. On a paid plan, budget models like GPT-5.4 Nano, Gemini 3.1 Flash Lite, and Claude 4.5 Haiku cost just as little; their exact token rates vary.

Understanding the Ratings

Every model has three ratings on a 1–5 scale:

• Creativity — the quality and originality of the prose. Higher means more varied vocabulary, more surprising metaphors, and more distinctive voice.
• Consistency — how closely the model follows your instructions, style settings, and established context. A 5/5 means it does exactly what you ask, every time.
• Freedom — how freely the model handles mature content, violence, and sensitive themes.

The Freedom Rating

This rating is especially important for genre fiction writers. Here's what each level means in practice:

• 5/5 (Unrestricted) — no content restrictions whatsoever. Handles explicit romance, graphic violence, horror, and taboo themes without refusing or softening. Only Grok 4.3 has this rating in the current lineup.
• 3/5 (Moderate) — handles most content but may soften explicit scenes or add content warnings. Fine for most fiction, but may frustrate writers of very dark or explicit material.
• 2/5 (Conservative) — will decline or significantly tone down explicit content. Best for clean fiction, young adult, or writers who prefer content guardrails.

Switching Models

You can switch models at any time using the model selector in the toolbar. Your choice persists for the current session. A few things to keep in mind:

• Switching models does not affect already-generated text — only future generations.
• Writing styles can specify a preferred model, but you can override this in the Advanced Settings panel.
• Different models cost different amounts. Check the cost indicator (1–5 cost) before generating if you're watching your token balance.

Want to understand how model costs translate to your token balance? Read The Token Economy.