The best AI for writing dialogue
Live data · ranked by Dialogue · measured as of July 23, 2026
Good dialogue is the fastest way to tell a strong model from a weak one: distinct voices, subtext under the words, exchanges that carry the scene instead of narrating it. This ranking scores the frontier models on exactly that — and the order is not the one you would guess from overall craft.
Key takeaways
- Models are ranked by measured dialogue craft — voice, subtext, and exchanges that move a scene — on real novel prose.
- The GPT-5.6 models (Sol and Terra) and Fable 5 lead dialogue.
- Overall craft does not predict dialogue: some models strong elsewhere write notably flat conversations.
- The ranking is live and updates as more prose is measured.
The ranking
Ranked by Dialogue
Ranked highest-to-lowest by measured dialogue craft. The score is each model’s confidence-adjusted value on the dialogue axis, 0–100.
- GPT-5.6 SolOpenAI83
Leads dialogue. Characters sound like themselves and exchanges have snap — pair that with its top action scores and it is a natural fit for banter-heavy, plot-driven fiction.
- GPT-5.6 TerraOpenAI82
Nearly matches Sol on dialogue at lower cost — the value pick for talk-heavy scenes.
- Fable 5Anthropic80
Anthropic’s storytelling model gives dialogue real personality and warmth.
- Gemini 3.5 FlashGoogle79
- Claude Sonnet 5Anthropic76
- Gemini 3.1 ProGoogle72
- Claude Opus 4.8Anthropic · Novelmint Lead71
The benchmark’s top overall model, but dialogue is its softest craft axis — it leans interior. On Novelmint, talk-heavy beats can deviate away from it while it keeps the narration.
- GPT-4.1OpenAI64
- Grok 4.3xAI61
- Claude Haiku 4.5Anthropic60
- Grok 4.5xAI49
The weakest dialogue in the field — functional rather than characterful. Use it for what it is good at (explicit content, action), not conversation.
How this ranking is made
Scored on how people talk
Each model writes the same briefed beats, and a blind judge panel rates the dialogue — whether characters sound like distinct people, whether subtext survives, whether the talk moves the scene — without knowing the author. Thin samples are shrunk toward a neutral baseline. This page orders the models by that one axis.
See the full nine-axis benchmark and methodologyQuestions
Frequently asked
- Which AI writes the best dialogue?
- On measured dialogue craft, GPT-5.6 Sol leads the Novelmint benchmark, with GPT-5.6 Terra and Fable 5 close behind. The live ranking on this page shows the current order.
- Is Claude good at dialogue?
- Claude Opus 4.8 is the strongest model overall but dialogue is its weakest craft axis — it favours interiority. Claude’s Fable 5 writes more characterful dialogue. For the sharpest exchanges, the GPT-5.6 models lead.
- Why does overall craft not predict dialogue?
- Dialogue is its own skill. A model can write gorgeous narration and still write flat conversations, which is exactly why the benchmark scores dialogue separately and Novelmint routes talk-heavy beats to the models that win it.
What this page does not claim
- This ranks measured dialogue craft on fiction beats only.
- A dialogue lead does not make a model the best overall — check the axes your book leans on.
- Model names are trademarks of their respective owners; this is an independent measurement.
관련
AI fiction model benchmark
How the frontier models actually write fiction — measured across nine axes.
Best AI for action scenes
Which AI keeps a fight or chase clear and grounded instead of vague. Ranked on physicality.
Best AI for conflict scenes
Which AI keeps an argument or standoff taut and rising. Ranked on measured conflict.
Best AI for emotional scenes
Which AI makes a feeling-led scene land instead of just describing it. Ranked on emotion.
Give your dialogue scenes to the model that writes them best.
Novelmint can route dialogue-heavy beats automatically. Your first chapter is free.