How Gemini 3.1 Pro writes fiction
Gemini 3.1 Pro is the workhorse of the set: measured over the largest sample of any model here, it is dependable, well-paced, and clear with physical action, and it handles mature and explicit content more openly than the GPT and Claude flagships. What it gives up is at the top of the range — its literary flair and emotional peak sit below the leaders. That makes it a steady, versatile choice rather than a specialist. A note on naming below.
Measured profile
Gemini 3.1 Pro
74
Overall craft
Craft — how well it writes
- Literary66
- Dialogue72
- Fidelity77
- Pacing80
Content — what it writes well
- Emotion73
- Physicality80
- Conflict77
- Romance74
- Eroticism77
Scores are confidence-adjusted, 0–100. * marks a provisional axis (fewer than 6 samples). Measured on gemini-3.1-pro-preview. See the full benchmark and methodology.
Key takeaways
Key takeaways
- Gemini 3.1 Pro is measured over the largest sample of any model in the benchmark, so its scores carry high confidence.
- Its strengths are reliability, pacing, and clear physical action — a dependable, versatile profile rather than a spiky one.
- It handles mature and explicit content markedly more openly than the GPT-5.6 and Claude flagships, which mostly decline.
- Its softer axes are literary flair and emotional peak, which sit below the field leaders.
- Naming: this is the specific version Novelmint measured (Gemini 3.1 Pro), not necessarily Google’s current “latest” — the two can differ, so the benchmark names the snapshot it tested.
How it writes
The most-measured, most dependable model here
Gemini 3.1 Pro has by far the largest sample behind it, so its shape is the best-established in the benchmark. And that shape is steadiness: solid fidelity, strong pacing, and clear physical action, with no alarming hole. It is the model least likely to surprise you — in either direction — which for long-form drafting is a genuine virtue.
Open where others decline
Unusually for a mainstream flagship, it handles mature and explicit content with relative willingness — it scores among the higher models on that axis, where the GPT-5.6 family and the Claude flagships mostly refuse or turn clinical. If a book needs a capable model that will actually engage with adult scenes, Gemini 3.1 Pro is a realistic option short of a dedicated permissive model.
Steady, not lyrical
The trade for that dependability is at the top of the craft range: its literary flair and its emotional peak sit below the leaders. Its prose is clean and competent rather than luminous, and its most feeling-heavy scenes land solidly rather than devastatingly. For a book chasing lyric intensity on every page, a higher-literary model leads; for steady, versatile output, this is a strong pick.
A word on which Gemini this is
Novelmint measured a specific version — Gemini 3.1 Pro — and the scores on this page describe that snapshot. Provider aliases move, so “our latest” and Google’s current “latest” are not always the same model; rather than imply otherwise, the benchmark names the exact version it tested. When we re-measure a newer version, its scores are taken fresh rather than blended with this one.
What the scores mean for a book
Lean on it for
- Dependable, well-paced drafting across a whole book.
- Clear physical action and scene mechanics.
- Mature or explicit scenes where you want a capable mainstream model that will engage.
Route elsewhere for
- The most lyric, literary passages, where Opus and the GPT-5.6 flagships lead.
- The most emotionally intense beats, where higher-emotion models have the edge.
In Novelmint
How Novelmint uses Gemini 3.1 Pro
Gemini 3.1 Pro is one of the default deviation-pool models: dependable and versatile, it is a safe alternate the router can lean on across a range of beats, and its openness on mature content makes it a candidate where flagships decline. Add it to your pool or set it as Lead for a steady build. Its ranking comes from the same live grid this page renders — for the exact version named above.
Questions
Frequently asked
- Is Gemini good at writing fiction?
- The version Novelmint measured, Gemini 3.1 Pro, is a dependable, well-paced model with clear physical action and unusually open handling of mature content — measured over the largest sample in the benchmark. It is softer than the leaders on literary flair and emotional intensity.
- Which version of Gemini is this benchmark measuring?
- Gemini 3.1 Pro specifically. Provider “latest” aliases move over time, so the version Novelmint currently measures and Google’s newest release can differ; the benchmark names the exact snapshot it tested rather than a generic “Gemini Pro”. Newer versions are re-measured from scratch, not blended with older ones.
- Can Gemini write mature or explicit scenes?
- More readily than most mainstream flagships. Gemini 3.1 Pro scores among the higher models on explicit content, where the GPT-5.6 family and Claude flagships largely decline. For the most explicit material a dedicated permissive model still leads, but Gemini is a capable middle option.
What this page does not claim
- These scores describe Gemini 3.1 Pro’s prose on fiction beats only, measured against Novelmint’s judged set — not an official Google rating, and nothing about reasoning or other capabilities.
- The scores describe the specific measured version and may not reflect a newer Gemini release; a newer version would be re-measured separately.
- Gemini is a trademark of Google; this is an independent measurement, not an endorsement.
Relacionado
AI fiction model benchmark
How the frontier models actually write fiction — measured across nine axes.
La mejor IA para escenas de acción
Qué IA mantiene una pelea o persecución clara y concreta en lugar de vaga. Clasificada por fisicalidad.
La mejor IA para escenas de conflicto
Qué IA mantiene una discusión o enfrentamiento tenso y en ascenso. Clasificada por conflicto medido.
La mejor IA para escenas emotivas
Qué IA logra que una escena emocional impacte en lugar de solo describirla. Clasificada por emoción.
A steady hand across the whole book.
Add Gemini to your build pool as a dependable alternate. Your first chapter is free.