Plans and Credits

Last updated: 2026-11-03. In short: a monthly plan grants Credits, every call deducts Credits at the model's price, your balance is always visible, and when it runs out you choose whether to buy a top-up pack or upgrade.

This page explains how DimAgent charges, what Credits are, which models each plan opens up, and what they cost. The plan, model pricing, and search allowance tables below come straight from the backend configuration — after a price change or a model being added or retired, reload this page and you see the new values. Prices in the estimates and discount explanations are a snapshot as of 2026-10-15.

Your own balance, effective prices, and usage live in the account center; this page only covers the rules.

How Billing Works

Subscribing to a plan grants a fixed amount of Credits every term;
Every conversation or task is converted into Credits from the token usage of the model you used;
Deduction records and the remaining balance are always visible in the account center;
You are warned before the balance runs out, and nothing is charged automatically afterwards: renewing and buying top-up packs are your decision.

Read on for the details.

Pick a Plan

A few things to note:

Higher tiers give more Credits per unit of money, which suits heavy use;
Lite+ is the larger Lite: twice the Credits plus another 200 Credits, and twice the granted search calls, with identical discounts and model access — for when Lite is not enough but Pro is more than you need;
Max+ is the larger Max: twice the entitlements, priced below buying two Max plans separately; identical discounts and model access;
Every plan renews manually. Auto-renewal is off by default, so a plan simply stops at the end of its term;
You can buy a top-up pack whenever you need more Credits — see below.

Nano

The entry-level tier: a small set of fast models — the exact list and prices are in "Models and Prices" below;
After it expires you can buy it again, or upgrade to a higher tier at any time;
Some of these models use flat pricing, others use peak / off-peak pricing — see below.

Models and Prices

Plans open up different models. Newly purchased or renewed plans bill glm-5.3 at 20% off all day; other models use standard or time-based pricing. Pick a plan to see its table:

About this table:

The prices shown are the final values used at deduction time — no further discount is applied on top;
The former night discount is for existing plans only: the promotion ended on 2026-10-15, and plans purchased before that date keep their night discount until the current term expires. See the GLM-5.3 section below; your current effective prices are in the account center;
A peak window belongs to the model's own pricing: the model uses a different price during those hours and returns to its base price outside them;
Plan discounts and model time-based pricing are separate rules; the table lists the windows returned by the backend on separate rows;
Windows are resolved against the Beijing time at which the request starts, hour by hour, start inclusive and end exclusive: a request sent at 11:59 is billed entirely at the earlier window's price;
Image models are charged per call (one call is one image) and do not use token pricing;
Model prices are reviewed monthly; changes are announced a month in advance.

Using Dim through magpie

After installing the magpie-dim-oauth plugin in the magpie client and signing in to your Dim account, model calls are billed against a separate price list. The table below shows the actual per-token prices for the public-test plans (Nano, Lite, Lite+, Pro, Max, Max+) when called through magpie, in Credits / 1M tokens. "—" means that model's cache-write price is 0 — no Credits are deducted separately.

Model	Required plan	Fresh input	Cache hit	Cache write	Output
`deepseek-v4.1-flash` (off-peak)	Nano and above	112	2.16	—	440
`deepseek-v4.1-flash` (peak)	Nano and above	224	4.32	—	880
`glm-5.3-flash`	Nano and above	110	32	—	380
`mimo-v2.6-flash`	Nano and above	140	2.8	—	280
`mimo-v2.6-pro`	Nano and above	435	3.6	—	870
`qwen3.8-27b`	Nano and above	70	10	—	140
`qwen3.8-flash`	Nano and above	150	16	200	470
`qwen3.8-max`	Nano and above	2,000	250	2,500	6,000
`seed-2.0-mini`	Nano and above	27.40	5.48	—	273.97
`glm-5.3`	Lite and above	1,200	260	—	4,200
`kimi-k3`	Lite and above	3,000	300	—	15,000
`kimi-k3.1-pro`	Lite and above	2,000	200	2,000	10,000
`seed-2.1-pro`	Lite and above	880	88	—	4,450

"Nano and above" means Nano, Lite, Lite+, Pro, Max, Max+; "Lite and above" excludes Nano.

Per-call models charge a fixed number of Credits per call:

Model	Required plan	Credits / call
`qwen-image-3.0`	Nano and above	30
`seedream-5.0`	Lite and above	40.14

What a Credit Is

A Credit is DimAgent's universal billing unit. What a request costs is decided entirely by the price of the model and the tokens the request consumed — there are no hidden fees.

Credits expire with the plan's 30-day term; whatever is left does not roll over and cannot be given away;
Pick a tier that matches your monthly usage — a low tier is plenty if you use it lightly;
Credits are not money: they cannot be cashed out or transferred, and refunds are not calculated from them.

How Credits Are Spent

Every model call deducts Credits from the tokens it consumed and the model's price:

Models differ in price: small fast models are cheap, flagship models (such as kimi-k3) cost considerably more;
Output costs more than input: long answers burn faster;
Cache hits are cheaper: context you reuse hits the cache and costs far less than fresh input;
An ordinary conversation costs roughly 0.2–5 Credits: fast small models sit at the low end, flagships at the high end — it depends on the model.

How many requests does each plan cover? (estimate)

Estimated at ~2,000 input + 500 output tokens per typical request, no cache hits (not a promise of fixed counts; deepseek-v4.1-flash is split into peak / off-peak rows). The table is a snapshot of the prices on 2026-10-15; live prices are in "Models and Prices" above:

Model	Cost per request	Nano / 1,500	Lite / 11,000	Lite+ / 22,200	Pro / 48,000	Max / 100,000	Max+ / 200,000	Coding pack / 2,500
`seed-2.0-mini`	0.19	~7,800	~57,400	~116,800	~250,300	~521,400	~1,042,800	~13,000
`qwen3.8-27b`	0.21	~7,100	~52,400	~105,700	~228,600	~476,200	~952,400	~11,900
`glm-5.3-flash`	0.41	~3,660	~26,800	~54,100	~117,100	~243,900	~487,800	~6,100
`mimo-v2.6-flash`	0.42	~3,600	~26,200	~52,900	~114,300	~238,100	~476,200	~6,000
`qwen3.8-flash`	0.54	~2,800	~20,400	~41,100	~88,900	~185,200	~370,400	~4,600
`deepseek-v4.1-flash` (off-peak)	0.56	~2,700	~19,600	~39,600	~85,700	~178,600	~357,100	~4,500
`deepseek-v4.1-flash` (peak)	1.11	~1,350	~9,900	~20,000	~43,200	~90,100	~180,200	~2,250
`mimo-v2.6-pro`	1.31	~1,150	~8,400	~16,900	~36,600	~76,300	~152,700	~1,900
`glm-5.3` (all-day 20% off)	3.60	not included	~3,100	~6,200	~13,300	~27,800	~55,600	~690
`seed-2.1-pro`	3.99	not included	~2,800	~5,600	~12,000	~25,100	~50,200	~630
`qwen3.8-max`	7.00	~210	~1,600	~3,200	~6,900	~14,300	~28,600	~360
`kimi-k3`	13.50	not included	~820	~1,640	~3,600	~7,400	~14,800	~190

(In Credits; actual counts vary with context length, tool calls, and cache hits. deepseek-v4.1-flash is billed by the window in effect when the request starts; glm-5.3 in the table is already at the all-day 20% off.)

Web Search

Every plan grants search calls per 30-day term, and the search itself costs no Credits:

Rules:

Each search consumes one call from the allowance;
The granted allowance resets every 30-day term and does not roll over; once used up, it stops;
Only searches that return successfully and are billed are counted; failed retries are not counted twice;
Once the free allowance is used up, searches draw from a Web Search top-up pack first; when both are exhausted, search stops for the rest of the term;
Feeding the search results to a model still costs Credits at that model's price.

Image Generation

Plans that include an image model can generate images directly with Seedream 5.0 at 40.137 Credits per image, deducted from the plan's Credits. Only successful outputs are billed; an image blocked by review is not.

With a Seedream top-up pack, generation draws from the pack first and leaves your Credits untouched.

Features That Stretch Your Credits

GLM-5.3: 20% off all day

glm-5.3 is not a separate plan — it is a model included in the regular plans. Newly purchased or renewed plans bill glm-5.3 at 20% off the standard price all day, with no separate night window.

GLM-5.3 price table (Credits / 1M tokens):

Tier	Discount	Input	Cache hit	Output
Standard	—	1,200	260	4,200
All-day 20% off	0.8×	960	208	3,360

The night promotion ended on 2026-10-15; newly purchased or renewed plans use the all-day 20% discount;
Plans purchased before 2026-10-15 keep their night discounts until the current term expires (original multipliers from 20:00 to 08:00 Beijing time: Lite / Lite+ 0.5×, Pro 0.3×, Max / Max+ 0.1×, calculated directly from the standard price without stacking discounts);
The discount applies to glm-5.3 only; other models use standard or time-based pricing;
Model prices are reviewed monthly; changes are announced a month in advance;
The Coding top-up pack bills glm-5.3 at 20% off all day.

DeepSeek peak / off-peak pricing

deepseek-v4.1-flash uses peak / off-peak pricing: peak hours are Monday to Friday 09:00–12:00 and 14:00–18:00 Beijing time (excluding Chinese public holidays); all other hours are off-peak (including weekends and Chinese public holidays).

deepseek-v4.1-flash corresponds to DeepSeek's official deepseek-flash (DeepSeek-V4.1-Flash, with vision and thinking modes); it replaces the earlier deepseek-v4-flash and deepseek-v4-flash-vision-exp, merging and upgrading their capabilities.

deepseek-v4.1-flash (per 1M tokens):

Item	Peak	Off-peak
Input · cache miss	280 Credits	140 Credits
Input · cache hit	5.4 Credits	2.7 Credits
Output	1,100 Credits	550 Credits

The window is resolved against the Beijing time at which the request starts, hour by hour, start inclusive and end exclusive;
For example, a request sent at 11:59 is billed entirely at the peak price; from 12:00 it switches to off-peak;
For long, repeated conversations, enable context caching: a cache hit costs about 1/52 of a miss on deepseek-v4.1-flash.

Top-Up Packs: A Little More When You Run Short

Pack	Price	Contents	Validity
Coding pack	¥20	2,500 Credits (`glm-5.3` at the all-day 20% off)	7 days
Web Search pack	¥20	2,000 searches	7 days
Seedream image pack	¥20	60 images	7 days

When to use one: near the end of a term, or a task that eats a lot at once — top up any time.

Worth knowing:

A pack can only be bought while the account has an active plan; the three kinds can be bought separately or held at the same time;
A pack has its own lifetime: 7 days from the moment of purchase, unaffected by renewing or upgrading the main plan;
Whatever is left when a pack expires is lost immediately — it is not frozen and does not roll over;
Packs never renew automatically; the Coding pack bills glm-5.3 at the all-day 20% off and other models at standard or time-based prices.

Invite Friends, Save Together

Want a friend on board? Share your personal invite link from the "Referral" page in the client:

The inviter needs an active plan;
The invitee must be a brand-new account;
Once the invitee buys any plan, both sides receive a reward of 20% of that plan's Credits. If your friend buys Lite (11,000 Credits), you and your friend each get 2,200 Credits; buying Lite+ (22,200 Credits) means 4,440 Credits each;
The reward follows the plan the invitee actually bought — more invites, more rewards;
Reward Credits expire with the main plan's 30-day term;
Each new account can take part once.

FAQ

Q: What happens when my plan expires? Usage stops and unused Credits are cleared. You can renew manually before it expires (the new term starts when the current one ends) or resubscribe afterwards (the term starts at payment).

Q: Will I be charged automatically? No. Auto-renewal and automatic top-up purchases are off by default, both enabling them and every charge require an explicit confirmation, and running out of Credits never triggers a recharge.

Q: What does one conversation actually cost? It depends on the model and the tokens. Take deepseek-v4.1-flash: a conversation of about 5,000 input + 1,500 output tokens costs about 1.5 Credits off-peak, about 3 Credits at peak; with seed-2.0-mini it is about 0.6 Credits — a cheap model barely dents your balance even after a full day.

Q: Can I upgrade mid-term? Yes. Upgrades take effect immediately, the unused value of the current term is credited against the new one, and a fresh 30-day term starts on the upgrade day. The target tier's full Credits are granted and the old tier's remaining Credits are not kept. Downgrades are not supported.

Q: Why can't I transfer Credits to a friend? Plans are billed and rate-limited per account, so Credits cannot be transferred, cashed out, or pooled. It is simpler for each person to buy their own.

Q: Are there usage limits? To keep the service stable for everyone, concentrated use is subject to a window cap. Reaching it pauses spending and the client shows when it recovers; a top-up pack raises the cap (it does not reset what you have used). This is quota control, not an upsell. Model discounts do not change these caps — they just make Credits last longer, so the cap arrives later too.

Q: How is a pack's 7-day validity counted? From the moment of purchase — 7 days, unaffected by renewing or upgrading the main plan; when it expires, whatever is left is lost immediately and needs to be bought again.

Q: Where are the invoice and refund policies? Payment, invoicing, and refund terms follow the site's Terms of Service and your order page.

Our Commitments

Transparent pricing: every plan is billed against the same model prices — the price never depends on which plan you are on;
Advance notice: price or entitlement changes are announced on the product pages, or by other reasonable means, before they take effect.

This page only describes how plans and Credits work. The full rights and obligations follow the site's Terms of Service; final interpretation rests with Shanghai Fazhi Technology Co., Ltd. (Fazhi).