AI Models

Qritor integrates a curated selection of top-tier AI language models, from economical to flagship, to suit different creative needs and budgets.

Model Tiers

Qritor organizes AI models into four tiers. Higher tiers offer stronger capabilities at a correspondingly higher credit cost.

TierDescription
EconomyBest value — daily writing, quick tests, short chats
StandardBalanced — mid-to-long-form creative writing
AdvancedMassive context — ultra-long novels, full-book style control
FlagshipTop capability — premium, high-quality creative work

---

Economy

DeepSeek (Default)

DeepSeek V3.2 Thinking is Qritor's default model — a leading Chinese-developed reasoning model. Built-in Chain-of-Thought capability delivers excellent reasoning for Chinese creative writing at the best value.

  • Strengths: Chain-of-Thought reasoning, Chinese-optimized, fast, creative
  • Best for: Daily continuation, plot brainstorming, character dialogue
  • Context window: 131K tokens
  • Rate: 0.4x / 0.6x

GLM-4.7 Flash

A lightweight, ultra-fast model from Zhipu AI. Ideal for tasks where instant response matters most.

  • Strengths: Ultra-fast response, lightweight, Chinese-friendly
  • Best for: Quick tests, instant inspiration, short conversations
  • Context window: 128K tokens
  • Rate: 0.3x / 0.5x

MiniMax M2.5

MiniMax's narrative-focused Chinese model. Excels at maintaining long-form narrative coherence.

  • Strengths: Long context, narrative consistency, Chinese-optimized
  • Best for: Long chapter continuation, narrative cohesion
  • Context window: 200K tokens
  • Rate: 0.5x / 1.4x

---

Standard

Gemini 3 Flash

Google's fast large-context model with 200K tokens context. Rapid streaming output with strong long-document comprehension.

  • Strengths: Large context, fast streaming, multilingual
  • Best for: Full-book reference continuation, long chapter generation
  • Context window: 200K tokens
  • Rate: 0.7x / 3.5x

Kimi K2.5

Moonshot AI's creative dialogue model. Deep understanding of Chinese nuance with outstanding character portrayal.

  • Strengths: Natural Chinese dialogue, character portrayal, creative
  • Best for: Character design, dialogue generation, emotional writing
  • Context window: 131K tokens
  • Rate: 0.8x / 3x

GLM-5

Zhipu AI's Chinese narrative flagship. Consistent style and high output quality, ideal when stylistic stability matters.

  • Strengths: Stable style, fluent Chinese narrative, high output quality
  • Best for: Formal chapter writing, style-consistent scenes
  • Context window: 128K tokens
  • Rate: 1x / 3.2x

---

Advanced

Gemini 3.1 Pro

Google's latest flagship model with an extraordinary 1 million token context window — fit an entire long novel into a single conversation. Stable style with exceptional creative narrative.

  • Strengths: 1M token context, stable style, creative narrative
  • Best for: Million-word novels with full-book reference, global style consistency
  • Context window: 1,000,000 tokens
  • Rate: 2x / 12x

---

Flagship

Claude Sonnet 4.6

Anthropic's flagship mid-tier model — a top performer for creative writing. Nuanced character psychology, rich emotional depth, and precise style control make it the go-to for quality writing.

  • Strengths: Top creative writing, deep character portrayal, precise style control
  • Best for: Premium chapters, complex character design, high-quality continuation
  • Context window: 200K tokens
  • Rate: 3x / 15x

Claude Opus 4.6

Anthropic's most powerful model — industry-leading across the board. Unmatched narrative depth, character psychology, and stylistic mastery for writers who demand the best.

  • Strengths: All-around strongest, narrative depth, character psychology, exceptional style
  • Best for: High-demand premium writing, complex plots, literary-quality prose
  • Context window: 200K tokens
  • Rate: 5x / 25x

---

How to Choose

ScenarioRecommended Model
Daily writing, best valueDeepSeek (default)
Quick tests & instant inspirationGLM-4.7 Flash
Long-form narrative consistencyMiniMax M2.5
Character design & dialogueKimi K2.5
Consistent style writingGLM-5
Million-word full-book referenceGemini 3.1 Pro
High-quality premium chaptersClaude Sonnet 4.6
Literary excellenceClaude Opus 4.6

About Rates

The input rate / output rate shown next to each model in the model selector reflects the relative credit cost — a higher rate means more credits consumed for the same amount of text.

---

When You Encounter Network Errors

If an AI call fails, an error message appears in the chat. Common errors and how to handle them:

Error MessageCauseWhat to Do
Network connection failed, please check your network and retryLocal network issue or server unreachableCheck your connection, or switch to another model
Network connection timed out, please check your network and retryRequest timed out — model responding slowlyWait and retry, or switch to a faster model
AI service temporarily unavailable, please try again laterProvider server outageSwitch to a model from a different provider
The selected model is unavailable. Please switch to another modelModel has been taken offline or suspendedSelect a different model in the model selector

How to Switch Models

After an error, click the model selector to the left of the AI assistant input box and select a different model to continue writing. Your selection is saved automatically and will persist on next launch.

Backup Model Suggestions

Each provider operates independently — an outage from one provider does not affect others. When errors occur, switch to a model from a different provider:

  • DeepSeek / GLM unavailable: Switch to Gemini 3 Flash or Kimi K2.5
  • Gemini unavailable: Switch to DeepSeek or the Claude series
  • Claude unavailable: Switch to GLM-5 or Gemini 3.1 Pro