Quick Answer (checked 9 October 2026)
No AI company publishes a fiction ranking, so there is no official "best model for novel writing". On what they do publish: Claude Opus 5.5 is Anthropic's default pick, and Anthropic says it writes more clearly and naturally than earlier models. GPT-6 Astra is OpenAI's most capable model and the one it lists for document creation. Gemini 3.1 Pro is Google's top model in the Gemini app, from the $4.99 Plus plan. All three hold a whole novel in about 1 million tokens of context. Free plans get Claude Sonnet, GPT-6 Luna or Gemini 3.5 Flash-Lite. To write a whole novel with an outline, chapters in order, a story bible, editing and export in one place, Inkfluence AI is built for that job.
Why This Matters
There is no single best model for novel writing. The right pick depends on your plan, your budget and the job in front of you.
The three big AI companies replaced their lineups between spring and autumn 2026: GPT-5 and GPT-4o gave way to the GPT-6 family, Claude Opus 4.7 and Sonnet 4.5 to the 5.5 models, and Gemini 2.5 Pro to the Gemini 3 series. Much of the "best AI for novels" advice online still names the old models.
This guide sets the current models side by side on facts each vendor publishes: which plan includes the model, how much text it holds, how long one reply can be, what it costs, and what its maker says it is for. Where a vendor makes no claim about fiction or prose, we say so. The picks for outlining, drafting and editing are general writing advice, labelled as such.
Updated 9 October 2026 for the current models. Every vendor fact below was checked on the vendor's own pages that day and is listed in the Sources section. We have not run our own prose tests for this version, so there is no fiction leaderboard here.
Models covered: Claude Fable 5.1, Opus 5.5, Sonnet 5.5 and Haiku 5.5; GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna; Gemini 3.1 Pro, 3.8 Flash and 3.5 Flash-Lite. We come back to the tool layer at the end, because for a whole novel the workflow matters at least as much as the model.
What People Actually Write (First-Party Data)
Before the models, one piece of context most comparisons miss: what people actually write. Across the 79,000+ books created on our own platform, fiction is one of the two largest categories alongside practical how-to guides, spanning every genre from romance and fantasy to sci-fi and literary shorts. In the 30 days to 9 October 2026, general fiction made up 21.2% of the books written on Inkfluence and romance another 13.8%, against 16.8% for practical guides.
Length matters for model choice, so we also measured it. In a sample of 1,200 books finished on Inkfluence in those 30 days (picture and colouring books excluded), one fiction book in ten passed 32,000 words, and the longest ran to 114,790 words. On Anthropic's published conversion of roughly 555,000 words per million tokens, that longest book is about 207,000 tokens: a fifth of a 1M-token context window, but far beyond what a free chat window holds at once. At novel length, context, reply limits and workflow decide more than any single model's style.
The Current Models at a Glance
Every figure in this table comes from the vendor's own model, pricing or plan pages, checked on 9 October 2026. API prices are per million tokens (input / output). "Longest reply" is the maximum output per API call. The quoted description is the vendor's own one-liner.
| Model | What the vendor says it is for | Consumer app access | API price | Context | Longest reply |
|---|---|---|---|---|---|
| Claude Opus 5.5 | "For long-running agentic coding and knowledge work"; Anthropic's default pick | Claude Pro ($20/mo) and up | $4 / $20 | 1M | 128K |
| Claude Fable 5.1 | "For demanding reasoning and long-horizon agentic work" | Pro via usage credits; Max (from $100/mo) | $10 / $50 | 1M | 128K |
| Claude Sonnet 5.5 | "The best combination of speed and intelligence" | Free plan (Sonnet) and all paid plans | $2 / $10 | 1M | 128K |
| Claude Haiku 5.5 | "For high-volume, latency-sensitive tasks such as classification, extraction, and routing" | Free plan and all paid plans | $0.10 / $0.50 (prompts up to 100K) | 1M API, 500K in the app | 128K |
| GPT-6 Astra | "Our most capable model for the most demanding work" | ChatGPT Plus ($20/mo) in Work and Codex; Pro reasoning in Chat from $100/mo | $10 / $50 | 1.05M | 128K |
| GPT-6.1 Sol | "Near-Astra performance for complex work at a lower cost" | ChatGPT Plus and up | $2 / $10 | 1.05M | 128K |
| GPT-6 Luna | "Our most efficient model for focused, high-volume tasks" | ChatGPT Free and Go ($8/mo) | $0.10 / $0.50 | 1.05M | 128K |
| Gemini 3.1 Pro (Preview on the API) | "built for multimodal understanding, agentic capabilities, and vibe-coding" | Google AI Plus ($4.99/mo) and up | $2 / $12 (prompts up to 200K); $4 / $18 above | 1,048,576 | 65,536 |
| Gemini 3.8 Flash | "Our most intelligent Flash model, engineered for long-horizon software engineering, autonomous agents, and complex enterprise workflows" | Google AI Pro ($19.99/mo) and Ultra | $0.75 / $3.75 to 31 Dec 2026, then $1.50 / $7.50 | 1,048,576 | 65,536 |
| Gemini 3.5 Flash-Lite | "fastest, most cost-effective 3.5 model for high-throughput execution" | Free Gemini app | $0.30 / $2.50 | 1,048,576 | 65,536 |
US list prices for individual plans; they vary by country. App context can be smaller than the API: ChatGPT publishes 27K tokens (Free), 54K (Go, Plus) and 128K (Pro) for everyday chats, and 256K or 400K with reasoning. Claude's paid plans give 1M on Opus 5.5, Sonnet 5.5 and Fable 5.1, 500K on Haiku 5.5. Google states 1M for the Gemini app on AI Pro.
How We Compared Them
General benchmarks (reasoning, code, recall) say little about fiction, and no vendor publishes a fiction benchmark. So we compared only what can be checked, and say plainly where nothing can.
Vendor pages only
Every model fact comes from Anthropic, OpenAI or Google's own documentation, pricing and plan pages, checked on 9 October 2026.
Six things that decide fit
Consumer plan access, context window, longest reply, API price, the vendor's own description, and its published writing guidance.
Vendor writing claims, quoted
Writing claims are quoted exactly. As of 9 October 2026, none of the three vendors makes a fiction claim for any current model.
No prose scores
There is no leaderboard. The picks for outlining, drafting and editing are general writing advice built on published limits and prices.
Costs as a floor
Novel costs use list prices and stated assumptions; reasoning tokens and revisions push real runs above the floor.
Your own test decides prose
For voice and style, run your own scene brief on two or three models and read the results blind. Tools like Inkfluence AI add the outline, story bible and chapter order a single prompt leaves out.
To grade models or tools yourself, our framework for evaluating AI novel quality sets out seven things to check, from character voice to pacing, and a 30-minute recipe for running one brief across tools.
Claude Opus 5.5 and Fable 5.1: Anthropic's Top Models
Opus 5.5: released 22 September 2026, $4 / $20 per million tokens, 1M context, 128K-token replies. Fable 5.1: released 1 September 2026, $10 / $50, same limits.
Anthropic's own advice: "If you're unsure which model to use, start with Claude Opus 5.5 for most workloads." Its launch page says Opus 5.5 "communicates more naturally than prior models", that "early testers found its writing clearer and easier to follow", and that it "puts the most important information up front". Those are claims about clarity, not fiction. Anthropic makes no claim about dialogue, character voice or storytelling for any current model.
Fable 5.1 costs 2.5 times Opus per token and is Anthropic's slowest current model, positioned for "demanding reasoning and long-horizon agentic work". Anthropic publishes two notes on its prose: "In some cases its prose is denser than Claude Fable 5's, with longer sentences and fewer paragraph breaks", and "Claude Fable 5.1's writing has few stock phrases and little unexplained jargon". Its prompting guide even offers a fix for ornate style: "Please remove all mannered prose."
The hard numbers are generous on both: 1M tokens is roughly 555,000 words on Anthropic's tokenizer, so manuscript, outline and character notes fit at once, and a 128K-token reply is about 71,000 words (thinking tokens count toward it). On claude.ai, Opus needs Pro ($20 a month) and Fable on Pro runs on usage credits. Choosing between Claude models only? Our guide to the best Claude model for writing covers all four, plan by plan.
Where it fits for novelists (general advice)
Opus 5.5 is a sensible paid default for chapters you intend to keep and for line-editing passes. Fable 5.1 is hard to justify for drafting on price; it suits long structural work that must end as one finished document. Test both against a page of your own prose first.
Claude Sonnet 5.5 and Haiku 5.5: The Everyday Claude Models
Sonnet 5.5: released 28 September 2026, $2 / $10, 1M context. Haiku 5.5: released 7 October 2026, $0.10 / $0.50 for prompts up to 100K tokens.
Sonnet 5.5 is half the per-token price of Opus 5.5 and faster. Anthropic says it "writes more clearly than our previous generation of models", and its model-selection guide lists content creation under Sonnet 5.5. The free claude.ai plan includes Sonnet and Haiku, so most writers trying Claude for free are using Sonnet (Anthropic does not say which version).
If you still see Sonnet 4.5 recommended for fiction, note the dates: deprecated 30 September 2026, retiring 30 November 2026. Opus 4.7 is now a legacy model behind Opus 5.5.
Haiku 5.5 is the budget model. Anthropic makes no writing claims for it and frames it around "high-volume, cost-sensitive tasks". It is useful around a novel rather than for it: chapter summaries, character lists, blurb variations, name checks.
One change for API users: on Claude 4.7 and later, temperature and similar sampling parameters return an error, and Google has removed them from Gemini 3 and later. Style now comes from the prompt: a sample of your prose, point of view and tense, and plain instructions such as Anthropic's "Your response should be composed of smoothly flowing prose paragraphs."
Where it fits for novelists (general advice)
Sonnet 5.5 suits active drafting where you regenerate scenes often, at half the per-token cost of Opus. Keep Haiku 5.5 for bookkeeping. For the story-bible side, our story bible generator gives you a free template.
GPT-6 Astra and GPT-6.1 Sol: OpenAI's Top Models
GPT-6 Astra: $10 / $50, 1.05M context, 128K-token replies, knowledge cutoff 30 April 2026. GPT-6.1 Sol: $2 / $10, same limits.
OpenAI calls GPT-6 Astra "our most capable model for the most demanding work" and lists its uses as "complex reasoning, coding, computer use, research, and document creation". GPT-6.1 Sol "delivers near-Astra performance at a lower cost for complex coding, computer use, and professional work". Neither description mentions fiction or prose.
OpenAI does publish writing guidance for its newest models, and it is practical for novelists. It notes that "GPT-6 Astra tends to use lists, tables and Markdown to make responses scannable. If your application needs prose with less formatting, specify that preference." Its suggested prompt asks for "clear, concise paragraphs, each developing one main idea", and a second targets stock phrases such as "delve" and "foster". If GPT drafts read like a report, start there.
In ChatGPT, Plus ($20 a month) gets GPT-6 at Medium and High thinking (powered by GPT-6 Sol), GPT-6.1 Sol, and GPT-6 Astra in ChatGPT Work and Codex. GPT-6 Pro, powered by Astra, is in Chat from the $100 Pro plan. Free and Go get neither Sol nor Astra. On Plus, OpenAI publishes 54K tokens of context for everyday chats and 256K with reasoning, which OpenAI puts at roughly 40 and 320 pages of input.
Where it fits for novelists (general advice)
A reasoning model at higher effort earns its cost on structure: the outline, clue-and-payoff planning, and checking chapter plans against what came before. GPT-6.1 Sol costs the same per token as Claude Sonnet 5.5, a fair second opinion for drafting. Inkfluence's outline tool and mystery and thriller workflow handle the planning step without any model choice on your side.
Gemini 3.1 Pro and 3.8 Flash: Google's Current Models
Gemini 3.1 Pro: Preview on the API, $2 / $12 up to 200K-token prompts, 1,048,576-token context, 65,536-token replies. Gemini 3.8 Flash: September 2026, $0.75 / $3.75 until 31 December 2026, same limits.
Gemini's long-context lead from 2025 has closed: GPT-6 and current Claude models also list about a million tokens. The difference left is the reply limit: 65,536 tokens per Gemini reply, half the 128K of GPT-6 and Claude. That rarely matters chapter by chapter, but it does when one call must rewrite a long section.
Google's descriptions are technical: 3.1 Pro is "built for multimodal understanding, agentic capabilities, and vibe-coding", and 3.8 Flash is pitched at software engineering, agents and enterprise workflows. The one writing mention we found is in Google's 3.8 Flash notes, which list "writing drafts" among tasks suited to its low thinking-effort setting. Google makes no fiction or prose claim.
In the Gemini app, Google AI Plus ($4.99 a month in the US) includes 3.6 Flash and 3.1 Pro, AI Pro ($19.99) adds 3.8 Flash and a stated 1M-token window, and Ultra starts at $99.99. Plus is the cheapest way to reach a vendor's top model in a consumer app.
Where it fits for novelists (general advice)
Gemini 3.1 Pro on the Plus plan is a low-cost way to work with a whole manuscript in one place, for reading it back, finding loose threads or summarising earlier books in a series. Gemini 3.8 Flash suits fast first drafts. For series writers, see Inkfluence's workflow for series authors and multi-book series writer: the story bible carries continuity across books without feeding every prior word back into a single prompt.
GPT-6 Luna and Gemini 3.5 Flash-Lite: The Free and Low-Cost Options
GPT-6 Luna: $0.10 / $0.50, 1.05M context. Gemini 3.5 Flash-Lite: $0.30 / $2.50, 1,048,576 context. Claude Haiku 5.5: $0.10 / $0.50 for prompts up to 100K tokens.
These models run the free consumer tiers. ChatGPT Free and Go use the Luna models (the pricing table lists GPT-6 Luna; the help centre says GPT-5.6 Luna is becoming the Free and Go default). The free Gemini app uses 3.5 Flash-Lite. Claude's free plan is the exception: it includes Sonnet as well as Haiku.
No vendor positions these models for long-form prose. OpenAI calls Luna "our most efficient model for focused, high-volume tasks"; Google built 3.5 Flash-Lite for "high-throughput, low-cost execution for subagent tasks and document parsing". They suit brainstorming, short scenes and summaries. The bigger limit on a free tier is the window around the model: ChatGPT Free publishes a 27K-token context for everyday chats, which OpenAI puts at roughly 12 pages of input. That is a chapter or two, not a manuscript.
Where it fits for novelists (general advice)
Use free tiers to test a premise, rough out scenes and generate names or blurbs. For drafting chapters that need to remember the rest of the book, the free chat window runs out before the story does. Inkfluence's free plan (5 chapters to start plus 5 more every month, no credit card) writes chapters in order with the story so far in view.
Best Model by Genre
No vendor makes a genre claim, so this is general writing advice, not a test verdict: what each genre asks of a model, and where to start.
| Genre | What the job needs | Where to start |
|---|---|---|
| Literary fiction | Sentence-level control and many revision passes. | A top paid model (Opus 5.5, GPT-6 Astra, Gemini 3.1 Pro), judged against your own prose. |
| Romance | Long, dialogue-heavy chapters and an emotional arc that holds. See our AI romance novel writer. | A mid-priced model for drafts (Sonnet 5.5, GPT-6.1 Sol, Gemini 3.8 Flash). |
| Mystery / thriller | Clues planted early that pay off late. See our AI mystery thriller writer. | Higher reasoning effort for the outline, then any capable model. |
| Fantasy | Names, places and magic rules that stay fixed for hundreds of pages. | Any model, with a story bible in every chapter request. |
| Sci-fi (hard) | Internal logic of invented systems. | Higher reasoning effort for world rules; check the science yourself. |
| Sci-fi (literary) | Character-driven prose with a speculative frame. | As literary fiction: top model, your style sample, several passes. |
| Series book 4+ | Earlier books available for reference. | A 1M-token API window, or a running series bible instead of pasting every book. |
| Children's fiction | Short text at a set reading level. See our children's book guide. | Any model; spend your effort on the brief and the read-aloud test. |
Real Cost Per Novel
Token prices fell sharply in 2026, and most published cost comparisons are stale. Here is the raw API floor for a 70,000-word novel at list prices checked on 9 October 2026. Assumptions: 25 chapters, each request carrying about 9,000 words of outline, story bible and recent chapters; Anthropic's published rate of about 1.8 tokens per word, applied to every row (tokenizers differ, so treat non-Claude rows as approximate); no thinking tokens and no regenerations.
| Model | Input $/1M | Output $/1M | Per 70k novel (floor) |
|---|---|---|---|
| GPT-6 Astra | $10 | $50 | ~$10.35 |
| Claude Fable 5.1 | $10 | $50 | ~$10.35 |
| Claude Opus 5.5 | $4 | $20 | ~$4.14 |
| Gemini 3.1 Pro | $2 | $12 | ~$2.32 |
| Claude Sonnet 5.5 | $2 | $10 | ~$2.07 |
| GPT-6.1 Sol | $2 | $10 | ~$2.07 |
| Gemini 3.8 Flash (to 31 Dec 2026) | $0.75 | $3.75 | ~$0.78 |
| Gemini 3.5 Flash-Lite | $0.30 | $2.50 | ~$0.44 |
| GPT-6 Luna / Claude Haiku 5.5 | $0.10 | $0.50 | ~$0.10 |
| Inkfluence AI Premium (flat) | - | - | $19.99/mo, unlimited chapters |
The floor is about 126,000 output and 405,000 input tokens per novel. All three vendors bill reasoning or thinking tokens as output, and regenerations repeat the cost, so real runs cost more. Gemini 3.8 Flash rises to $1.50 / $7.50 on 1 January 2027 (about $1.55 per novel on the same assumptions).
The honest reading: raw tokens are no longer the expensive part of an AI novel. Even top models land around $10 for a first pass. What costs time is everything around the tokens: planning, keeping chapter 14 consistent with chapter 3, revising, and turning a manuscript into a cover, an EPUB and a print file. We worked through the full sums in our BYOK vs all-included cost breakdown.
Why Picking a Tool Beats Picking a Model
The table tempts the technical reader to build a personal stack: one model for the outline, another for drafting, a cheap one for summaries. It can work, but the model is the smallest part of the job, and most people stall on the parts that have nothing to do with which model they picked.
Writing a novel with raw models means building by hand what a book needs: an outline the chapters follow, a story bible of characters, places and events, chapter requests that carry the story so far, a way to catch a character who learns the same secret twice, an editing pass, and export to EPUB, print-ready PDF and DOCX. None of that is writing; it is project management.
This is the gap a tool fills, and why Inkfluence AI is our pick for writing a whole novel. You give it your premise; it builds an outline you can edit, then writes the chapters in order, each with the story so far and the story bible in view. The story bible updates as the book grows (who knows what, where everyone is, what has happened), so later chapters build on earlier ones instead of repeating them. An AI editing assistant works on any chapter, and the book exports to PDF, EPUB, DOCX and KDP print files, with a cover and audiobook narration in the same place. You never pick a model or manage a key.
The limits: the free plan gives you 5 chapters to start plus 5 more every month, with PDF export carrying a small Inkfluence credit. EPUB and DOCX export, the editing assistant and audiobooks start at Creator ($9.99 a month, 35 chapters); Premium ($19.99) has unlimited chapters. Length comes from the number of chapters, typically 1,000 to 2,000 words each.
For the platform-by-platform comparison see our ranked roundup of the best AI novel writers in 2026, our novel-writing software listicle with the per-novel cost table, and our deep dive on tools that finish 70,000+ word books. If you are weighing a chat assistant against a book tool, Inkfluence AI vs ChatGPT and Inkfluence AI vs Claude set them side by side.
The shortcut
Skip the stack-building. Try Inkfluence AI's novel writer free. Start with an outline, write your opening chapters in order with a story bible that keeps up, then export to PDF. The first 5 chapters are free, with 5 more every month, no API key and no credit card.
What Will Change This Year
Dated changes already announced by the vendors, worth planning around:
- Claude Sonnet 4.5 retires on 30 November 2026. Anything still built on it moves to a newer Claude model.
- Gemini 3.8 Flash doubles in price on 1 January 2027, to $1.50 / $7.50 per million tokens. Gemini 3.1 Pro is still labelled Preview on the API.
- GPT-6 is still rolling out in ChatGPT to eligible paid users in stages, per OpenAI's help centre, so the model behind a given plan may shift again.
- AI text watermarking. Anthropic marks text from its current models with a statistical watermark that "may persist through some editing". Store disclosure rules apply whatever the model; see our KDP AI disclosure guide.
If model choice is the deciding factor for you, the bring-your-own-key studios are covered in Inkfluence AI vs ProseEngine and Inkfluence AI vs Novelcrafter, and the choice between a built-in model and your own key is laid out in Sudowrite vs Novelcrafter.
FAQ
What is the best AI model for writing a novel in 2026?
Is Claude better than GPT-5 for writing fiction?
What is the cheapest AI model that can write a full novel?
Which AI model has the largest context window for writing long novels?
Does Claude Sonnet 4.5 actually write well enough to publish?
Should I use GPT-5 or Claude for outlining a novel?
Which AI model handles character continuity best across chapters?
What is the best free AI for writing a novel?
Are AI fiction benchmark scores reliable?
Which AI model should I use to write a romance novel?
Sources
Every vendor fact in this guide comes from the vendor's own pages, each checked on 9 October 2026. Prices and plan contents change often; check the vendor page before you buy.
Anthropic
- Models overview (lineup, descriptions, prices, context, output, words per token, default advice): platform.claude.com/docs/en/about-claude/models/overview, checked 9 Oct 2026
- Claude Fable 5.1 model page (released 1 September 2026, $10 / $50): platform.claude.com/docs/en/models/fable-5-1/overview, checked 9 Oct 2026
- What's new in Claude Fable 5.1 (denser prose): platform.claude.com/docs/en/models/fable-5-1/whats-new-fable-5-1, checked 9 Oct 2026
- Prompting Claude Fable 5.1 (stock phrases, mannered prose): platform.claude.com/docs/en/build-with-claude/prompt-engineering/prompting-claude-fable-5-1, checked 9 Oct 2026
- Prompting best practices (flowing prose paragraphs): platform.claude.com/docs/en/build-with-claude/prompt-engineering/claude-prompting-best-practices, checked 9 Oct 2026
- Choosing a model (content creation under Sonnet 5.5): platform.claude.com/docs/en/about-claude/models/choosing-a-model, checked 9 Oct 2026
- Pricing (Haiku 5.5 tiers, legacy prices): platform.claude.com/docs/en/about-claude/pricing, checked 9 Oct 2026
- Model deprecations (Sonnet 4.5 retirement, sampling parameters): platform.claude.com/docs/en/about-claude/model-deprecations, checked 9 Oct 2026
- Extended thinking (thinking tokens billed as output): platform.claude.com/docs/en/build-with-claude/thinking, checked 9 Oct 2026
- Claude plans and pricing: claude.com/pricing, checked 9 Oct 2026
- Introducing Claude Opus 5.5 (22 September 2026): anthropic.com/claude-opus-5-5, checked 9 Oct 2026
- Introducing Claude Sonnet 5.5 (28 September 2026): anthropic.com/claude-sonnet-5-5, checked 9 Oct 2026
- Introducing Claude Haiku 5.5 (7 October 2026): anthropic.com/claude-haiku-5-5, checked 9 Oct 2026
- Context window on paid plans: support.claude.com/en/articles/8606394, checked 9 Oct 2026
- Usage and length limits (summarising long chats): support.claude.com/en/articles/11647753, checked 9 Oct 2026
- How Claude marks AI-generated content (watermark): support.claude.com/en/articles/16266773, checked 9 Oct 2026
OpenAI
- Models (current flagship models, descriptions, prices, context, output): developers.openai.com/api/docs/models, checked 9 Oct 2026
- GPT-6 Astra model page (uses, knowledge cutoff): developers.openai.com/api/docs/models/gpt-6-astra, checked 9 Oct 2026
- GPT-6.1 Sol model page: developers.openai.com/api/docs/models/gpt-6.1-sol, checked 9 Oct 2026
- GPT-6 Luna model page: developers.openai.com/api/docs/models/gpt-6-luna, checked 9 Oct 2026
- Using the latest model (personality and writing style guidance): developers.openai.com/api/docs/guides/latest-model, checked 9 Oct 2026
- Reasoning models (reasoning tokens billed as output): developers.openai.com/api/docs/guides/reasoning, checked 9 Oct 2026
- ChatGPT pricing (plans, models per plan, context windows): chatgpt.com/pricing, checked 9 Oct 2026
- GPT-6 and other models in ChatGPT (help centre): help.openai.com/en/articles/20001354-gpt-6-and-other-models-in-chatgpt, checked 9 Oct 2026
- Introducing GPT-6.1 Sol: openai.com/index/introducing-gpt-6-1-sol/, checked 9 Oct 2026
- Gemini models: ai.google.dev/gemini-api/docs/models, checked 9 Oct 2026
- Gemini API pricing (descriptions, prices, thinking tokens billed as output): ai.google.dev/gemini-api/docs/pricing, checked 9 Oct 2026
- Gemini 3.1 Pro Preview model page: ai.google.dev/gemini-api/docs/models/gemini-3.1-pro-preview, checked 9 Oct 2026
- Gemini 3.8 Flash model page: ai.google.dev/gemini-api/docs/models/gemini-3.8-flash, checked 9 Oct 2026
- Gemini 3.5 Flash-Lite model page: ai.google.dev/gemini-api/docs/models/gemini-3.5-flash-lite, checked 9 Oct 2026
- What's new in Gemini 3.8 Flash (writing drafts, price change, sampling parameters): ai.google.dev/gemini-api/docs/latest-model, checked 9 Oct 2026
- Gemini app plans (US): gemini.google/us/subscriptions/, checked 9 Oct 2026
Inkfluence AI
- Content mix and book length: Inkfluence AI, books created and a sample of 1,200 books finished in the 30 days to 9 Oct 2026, excluding picture and colouring books; word counts summed from chapter text.
What to Read Next
Which Claude Model Is Best for Writing?
Opus 5.5, Sonnet 5.5, Haiku 5.5 and Fable 5.1 compared plan by plan, with the maths of fitting a whole book.
How to Evaluate AI Novel Writer Quality
A seven-dimension fiction rubric plus a 30-minute test recipe to grade any tool yourself.
BYOK vs All-Included: The Real Cost
Full-novel cost breakdown across model APIs, with the break-even point where flat-rate subscriptions win.
Best AI Tools for Long Novels in 2026
Tools (not models) that handle 70,000+ word books with continuity. Companion to this model-level comparison.
Best AI for Novel Continuity Checking
How tools layer continuity on top of base models to fix the character-drift problem.
Can AI Write a Novel? What Actually Works
The realistic AI novel workflow, what AI does well, where it falls short, which genres work best.
Best AI Novel Writer 2026 (Tools Roundup)
Eight novel-writing platforms ranked. Pairs with this model-level comparison to give you the full picture.
Founder, Inkfluence AI
Sam is the founder of Inkfluence AI. He built the platform to make book creation accessible to everyone - from first-time authors to seasoned publishers.
Helpful links
Ready to write your own book?
Write with AI, make it yours, and export PDF, EPUB and KDP-ready files from one project.
Keep reading
AI Writing NaNoWriMo Alternatives 2026: Every November Writing Challenge, and Which Ones Allow AI
NaNoWriMo closed in 2025, but November is still novel month. Here is every 2026 challenge that replaced it, with dates, goals, prizes, age limits and the one thing other lists leave out: what each challenge says about AI. Plus a word-count planner, a Preptober plan for the last few days of October, and two honest 30-day plans: one for the challenge, one for an AI-assisted November that ends with a published book.
AI Writing Prix Pascal: The New Literary Prize for Novels Written with AI, and How to Enter by 30 November 2026
France's new Prix Pascal is a literary prize that asks entrants to say how they used AI instead of hiding it. The rules in plain English, what 300,000 characters means in words and chapters, the technical file, and a five-week plan to submit a French novel on time.
AI Writing Sudowrite vs Novelcrafter (2026): Price, Built-in AI vs Your Own Key, Story Bible vs Codex
Sudowrite ($10 to $59 a month, AI built in) against Novelcrafter ($4 to $20 a month, AI from your own provider): price, how the AI is paid for, Story Bible against Codex, drafting, editing, export and series, read on both vendors' sites on 17 October 2026, with a picker.
Get ebook tips in your inbox
Join creators getting weekly strategies for writing, marketing, and selling ebooks.