Nowadays, everyone using an Agent can't do without a Coding Plan subscription package.
But there are too many packages on the market, and each provider's rules are different—quotas and available models all vary, making your head spin. Most importantly, the prices are not friendly, easily 199, 599... Every time you renew, your heart bleeds! (╥﹏╥)
Ipaid 260 yuan out of my own pocket, tested the current mainstream domestic7 Coding Plans, and made a ranking, striving to help friends understand the differences between each package and find one withlarge volume, low price, and fast speedfor high cost-effectiveness, stable and fast to use, so you no longer have to worry about falling into pitfalls and wasting money~
01. Coding Plan Review Comparison
I simply wrote a task to test:
Prompt: Choose one of 7 digital products including MacBook Neo, research its user reviews from the past 1 month, with data sources from the three major platforms Bilibili, Weibo, and Xiaohongshu, and finally generate an analysis report in Word document format, saved in the D:\Test folder.
Zhipu Coding Plan Lite Edition
Price: ¥49/month
Supported models: GLM-5.1, GLM-5-Turbo, GLM-4.7, GLM-4.6, GLM-4.5-Air
Quota: 80 prompts / 5 hours, 400 prompts / week
GLM-5.1 and GLM-5-Turbo are high-tier models; when called during peak hours (14:00-18:00), they consume quota at 3x, and at 2x during off-peak hours. Limited-time special offer before the end of June: only 1x consumption during off-peak hours.
Web search MCP / web page reading MCP / open-source repository MCP 100 times / month
The "Invite Friends, Group Buy Models" activity: inviting friends can earn 10% of the base bonus.
Model GLM-5-Turbo finished the task in 9 minutes; the generated document has excellent formatting and also includes table comparisons, with about 3,200 words. The actual test performance was very good.
Excellent in actual testing, but the package is only restocked in limited quantities every morning at 10:00, making it relatively hard to grab. Overall rating: top-tier.
Kimi Membership Andante Edition
Price: ¥49/month Supported models: Kimi K2.6, Kimi K2.5, Kimi K2 Quota: about 300-1200 requests every 5 hours, maximum concurrency 30.It should be noted that Kimi Code and the Kimi main site membership share the monthly total quota. If the web version uses up the quota, Kimi Code will also be restricted and you will need to wait for the quota to reset.
In actual testing, Kimi K2.6 took about 26 minutes to complete the task. The generated Word report was formatted, with very detailed information, totaling 3,139 words, and the sources were indicated at the end.
Overall performance is very good, but the Kimi K2.6 model simply took a relatively long time for the whole task. Overall rating: above average.
MiniMax Token Plan Starter Edition
Price: ¥29/month
Supported models: MiniMax-M2.7, M2.7-highspeed, TTS HD, Hailuo-2.3-Fast, Hailuo-2.3, Music-2.6, image-01
Quota: 600 model calls / 5 hours, 6000 model calls / week
Token Plan co-building invitation program: inviters get 10% of the actual paid amount back as platform vouchers, and invitees enjoy a 10% discount.
Using MiniMax-M2.7 for testing, it ran for 2 minutes, very fast, and generated a Word document of about 2,200 words. Plain text, but the content was all correct.
The price is relatively cheap, the speed is fast, actual test performance was average, and the quota is not particularly large. Overall rating: above average.
Xiaoma Computing Power Coding Plan Lite Edition
Price: ¥49/month
Supported models: GLM-5.1, MiniMax-M2.5
Quota: 1500 requests / 5 hours, 12000 requests / week, 22000 requests / month
MiniMax-M2.5 ran for 4 minutes and completed the task, meeting all the requirements, and the document formatting was good.
The price isn't very competitive, there are fewer models, it's quite fast, actual performance is average, you can switch between different invocation modes to handle different needs, and it's also very hard to get, so overall it gets an NPC.
iFlytek Xingchen Coding Plan Worry-Free Edition
Price: first purchase ¥3.90 / month, original ¥19 / month
Supported models: GLM-4.7-Flash, Qwen3.5-35B-A3B, Qwen3-Coder-Next-FP8
Quota: unlimited number of requests
Astron Coding Plan friend invitation event: the inviter and the invitee each receive a gift card equal to the payment amount, which can be used to offset platform model invocation costs.
Tested with Qwen3.5-35B-A3B, it ran for 9 minutes and ultimately produced reports in both Chinese and English; the Chinese version's title was in Martian text but the content was normal, about 2100 characters.
The original price of 19 is the lowest price in this review, and the first-purchase price of 3.9 is even better; you can switch models directly through the console, and there's no request limit.
Test performance is relatively average, there are few models and none of them are flagship models, so it's not practical enough, but it's cheap and unlimited, so it can be used for daily tasks; overall it gets an NPC.
Volcano Engine Ark Coding Plan Lite Edition
Price: ¥40 / month
Supported models: GLM-5.1, GLM-4.7, Kimi K2.6, Kimi K2.5, MiniMax-M2.7, DeepSeek-V3.2, Doubao-Seed-2.0-pro, Doubao-Seed-2.0-lite, Doubao-Seed-2.0-Code, Doubao-Seed-Code
Quota: 1200 requests / 5 hours, 9000 requests / week, 18000 requests / month.
For each friend you invite to place an order, you get a voucher worth 10% of their order amount, with no upper limit; friends can enjoy an extra 10% off their first subscription.
Using GLM-5.1, the Word document generated in 17 minutes actually had 5500 characters, the content was detailed, and the full text was divided into ten sections. The tables were detailed and clearly formatted, and it also summarized and analyzed the brand's public opinion; the final references included links, making it the best actual performance.
The quota is relatively large, there are many models, and the latest top domestic models from three companies can all be used; test performance is the best, so overall it gets a夯.
Baidu Qianfan Coding Plan Lite Edition
Price: ¥40 / month
Supported models: GLM-5, Kimi K2.5, MiniMax-M2.5, DeepSeek-V3.2, ERNIE-4.5-Turbo-20260402
Quota: 1200 requests / 5 hours, 9000 requests / week, 18000 requests / month
The GLM-5 model completed about 1900 characters in 6 minutes, meeting all requirements, though the content was relatively brief.
There are quite a few models, but they are not the strongest from each company at present; switching in the console is very convenient. The speed is okay, actual performance is average, there are no latest models, so overall it gets an NPC.
Finally, here's a full-dimension comparison chart to compare the strengths and weaknesses of each company's plans.
02. Some Sharing
For ordinary people and individual developers, choosing the right API plan can improve work efficiency by more than just a little.
Cost-effectiveness is not about being the cheapest, but about being the most suitable. Most of us don't need the top-tier plan; by clarifying our real needs, we can avoid paying for features we won't use. Refer to the rankings above and, based on your actual needs, just choose a plan that is sufficient.
Domestic manufacturers have gone crazy competing on price, constantly increasing discounts, offering discounts and vouchers, with all kinds of tricks emerging one after another, but when it comes to the actual experience, there is still a long way to go.
Due to limited time, there are actually still many Coding Plans on the market that I haven't had time to test one by one.