Model and pricing guide

Understand Low, Medium, High, Max, and Ultra across model groups, reasoning budget, pricing multipliers, and usage guidance.

Model 01

Five model levels

The Chron Engine selector is not only a speed control. It balances model group, reasoning budget, tool-use behavior, and pricing multiplier.

Chron Engine currently has five levels: Low, Medium, High, Max, and Ultra. They share the same interface, but send different model-routing, reasoning-budget, and behavior-strategy signals to the backend.

Low matches the previous default Chron Engine level and remains the recommended default for general use.After long-term use, we find its knowledge capability is enough for most daily scenarios, with a 1x pricing multiplier.

Low

The default level, priced at 1x. It uses the previous default Chron Engine model group and works well for most Q&A, organization, rewriting, summaries, and clear tasks.

Medium

Uses the same model group as Low and keeps the 1x multiplier. It enables reasoning for supported models in that group and allocates a higher budget, so token usage is usually higher.

High

Uses a stronger model group than Low and Medium, with a 3x multiplier. Best for complex analysis, multi-step planning, cross-source comparison, and higher-value questions.

Max

Uses the stronger model group at a 3x multiplier. It enables higher reasoning intensity and budget for supported models, plus a stricter, broader, more tool-active strategy.

Ultra

Off by default, with a 25x multiplier. It uses GPT-5.5 as the base model, xhigh reasoning effort, and the same prompt strategy as Max.

Use Low for general work. Move up to Medium, High, or Max only when the question is genuinely more complex or higher value.

Model 02

How Chron Engine uses models

Chron Engine model usage is dynamic and composite. A level is not a simple one-to-one binding to a fixed model.

We continuously adjust the model groups behind each level based on long-term testing, stability, answer quality, and cost performance. The product exposes Low, Medium, High, Max, and Ultra as operating levels, not as a permanent list of model names.

The current Chron Engine model pool includes DeepSeek 4 Pro, GLM-5.1, GLM-5.2, Qwen-series models, MiniMax 3, and related candidates. Different levels combine these models and tool paths into the execution strategy that fits the task.

A Chron Engine request is also not always completed by one model. In general, three primary models handle core judgment, reasoning, and generation, while smaller mixed models and tools support retrieval, file reading, search, code execution, and other task-specific steps.

Dynamic model groups

Model groups may change as our tests evolve. A level represents an ability and cost strategy, not a permanent model list.

Composite execution

Complex tasks may pass through understanding, retrieval, tool use, synthesis, and final response stages instead of a single model call.

Current model pool

Chron Engine currently uses models such as DeepSeek 4 Pro, GLM-5.1, GLM-5.2, Qwen-series models, and MiniMax 3, with adjustments based on testing.

Primary models, smaller models, and tools

Three primary models usually carry the core work, with smaller models and tools handling classification, extraction, search, and context loading.

So this page is best read as a level-strategy guide, not a fixed model roster. Actual consumption depends on the model, token, and tool path used for that request.

Model 03

How levels affect usage

The multiplier is only the first usage hint. Actual consumption also depends on reasoning budget, output length, tool-call count, and context size.

Medium and Low both use a 1x multiplier, but Medium enables a higher reasoning budget on supported models, so the same question may consume more tokens.

High and Max both use a 3x multiplier, with the main difference being reasoning strategy. Max thinks more rigorously and broadly, and it is more proactive with tools, so response time and token usage are usually higher than High.

01
Low
1x, recommended default, using the legacy default model group.
02
Medium
1x, same model group, with reasoning enabled for supported models.
03
High
3x, stronger model group for complex questions.
04
Max
3x, higher reasoning budget and more active tool strategy.
05
Ultra
25x, GPT-5.5 with xhigh reasoning effort, off by default.

Even under the same multiplier, enabling reasoning and increasing tool use can significantly raise actual token consumption.

Model 04

When to use Ultra

Ultra is off by default, and we do not recommend it for regular scenarios.

Ultra is expensive mainly because it uses GPT-5.5 as the base model and xhigh reasoning effort. It also follows the Max prompt strategy, so complex prompts can trigger many tool calls and extensive boundary exploration.

We tested 10 typical questions and found Ultra did not create a large gap from Max. In some cases, Max performed better. Unless you have an extreme requirement for the base model itself, prefer Max or a lower level.

Ultra must be enabled manually in settings, and your account needs enough credits. Final billing still depends on the actual model, token usage, and tool-related consumption recorded by the server.

Use Low for daily questions, Max for hard questions, and Ultra only for rare tasks with extreme base-model requirements.