19 Yuan/Month, GLM-5.2 + DeepSeek V4 Pro Unlimited – iFlytek's Move Crashes AI Programming Price Floor

A cup of milk tea costs 19 yuan. With that, you might get a full month of unlimited GLM-5.2 and DeepSeek V4 Pro.

Not a typo. The Astron Coding Plan revamped version on iFlytek Star MaaS platform starts at 19 yuan/month with unlimited requests for the Worry-Free plan. The Pro plan is 39 yuan, and the Efficient plan is 199 yuan. And—GLM-5.2 and DeepSeek V4 Pro are now available.

Key Points: Why These Two Models Matter

GLM-5.2: Domestic Flagship with 1M Context

GLM-5.2 is Zhipu's latest flagship, featuring a 1M context window—1 million tokens. What does that mean? You can feed it an entire medium-sized project's code, and it can still discuss architecture refactoring with you.

Under the MaaS platform's API pay-as-you-go model, GLM-5.2's input price is 8 yuan per million tokens, output 28 yuan per million tokens. Currently, there's a 60% discount across the board, making the post-discount input 4.8 yuan and output 16.8 yuan. But if you use the Coding Plan subscription, these costs are included—no separate charges.

DeepSeek V4 Pro: 1.6 Trillion Parameter Reasoning Beast

DeepSeek V4 Pro, with 1.6 trillion parameters in MoE architecture and 1M context, is one of the largest domestic models in terms of parameters. Its reasoning capabilities directly rival the top international tier.

API pay-as-you-go original price: input 12 yuan per million tokens, output 24 yuan per million tokens. After 60% discount: input 7.2 yuan, output 14.4 yuan. With the Coding Plan Efficient version, calls are included directly in the subscription.

These two models appearing together in a plan starting at 19 yuan/month—that's the real highlight.

Coding Plan Revamped: How to Choose the Three Tiers

The Worry-Free plan at 19 yuan with unlimited requests has almost no competitors at this price point. The Pro plan at 39 yuan offers about 1,200 requests per 5 hours, sufficient for daily coding. The Efficient plan at 199 yuan provides about 6,000 requests per 5 hours, suitable for developers who rely on AI programming.

Off-Peak Savings

From 22:00 to 08:00 the next day, all day on weekends, and all day on holidays, the billing factor is 0.8. That means your request quota automatically expands by 25% during off-peak hours—the Pro plan's 1,200 requests in 5 hours becomes 1,500, and the Efficient plan's 6,000 becomes 7,500.

Programmers are night owls anyway; this design is tailor-made for us.

What Models Are Included in the Plan?

Not just GLM-5.2 and DeepSeek V4 Pro. The Coding Plan covers a full list of domestic flagship models:

There are also a number of free models: Qwen3.5-35B, Qwen3.6-35B, DeepSeek-OCR, etc., with zero cost.

14+ paid models + 10+ free models, all accessible with one subscription.

Compatible with All Your Coding Tools

Coding Plan supports OpenAI protocol and Anthropic protocol access. This means:

• Cursor ✓
• Claude Code ✓
• OpenClaw ✓
• Continue ✓
• LobeChat ✓
• Any client that supports OpenAI API ✓

Configuration is as simple as changing the base_url and api_key—no other code changes needed. Using GLM-5.2 in Cursor to write code feels exactly the same as using GPT-4, but the price difference is dozens of times.

Invite Code: Double Cashback

iFlytek has a friend invitation activity with a simple mechanism: double rewards.

When you purchase using an invite code, both you and the inviter receive a gift card equal to the actual order amount. Buy the Pro plan at 39 yuan, both get 39 yuan gift cards. Buy the Efficient plan at 199 yuan, both get 199 yuan.

Gift cards can be used to offset API fees and model fine-tuning training costs. The invite code is valid indefinitely, with no monthly limit on invitations.

My invite code link:

👉 https://maas.xfyun.cn/packageSubscription?inviteCode=MAAS-E4248A94

Use this link to purchase, and we both get equal gift cards. It's like getting a 50% discount on your subscription.

Token Plan: Enterprise Choice

If your needs go beyond programming—research, content creation, data analysis—the Token Plan is more suitable. It charges per member, with each member receiving a fixed credit quota:

The Token Plan also supports GLM-5.2 and DeepSeek V4 Pro, and additionally includes a free AstronClaw Basic membership (worth 168 yuan) and 8 iFlytek features: ultra-realistic speech synthesis, one-click voice cloning, fast speech transcription, image/PDF OCR, receipt recognition, document proofreading, text translation, and image understanding.

Enterprises use the Token Plan, individual developers use the Coding Plan—both paths are open.

Let's Do the Math

Take the Pro plan at 39 yuan/month as an example:

• About 18,000 requests per month
• Access to 14+ flagship models including GLM-5.2 (1M context) and DeepSeek V4 Pro (1.6 trillion parameters)
• Off-peak quota automatically expands by 25%
• Purchase with invite code, both get 39 yuan gift cards (deductible from API fees)

39 yuan for a full month of domestic flagship model family bucket.

For comparison: ChatGPT Plus costs $20/month (about 145 yuan), only uses GPT-4o; Claude Pro costs $20/month, only uses Claude Sonnet. Meanwhile, the Coding Plan Pro at 39 yuan gives you GLM-5.2, DeepSeek V4 Pro, Spark X2, Kimi K2.6... switch freely.

It's not just a little cheaper—it's an order of magnitude cheaper.

Limited-Time 60% Discount: API Pay-As-You-Go Also on Sale

If you don't want to subscribe to a plan, you can still use the API pay-as-you-go model. Currently, there's a 60% discount across the board, valid until July 18:

GLM-5.2 input at 4.8 yuan per million tokens with 1M context—this price is almost unbeatable among 1M context models.

Final Thoughts

This move by iFlytek Star MaaS essentially does one thing: lowers the barrier to using domestic flagship models from "worrying about every token" to "unlimited monthly usage."

GLM-5.2's 1M context and DeepSeek V4 Pro's 1.6 trillion parameters—six months ago, a single call of these models would cost you a cup of coffee. Now it's 19 yuan per month, unlimited.

Invite link:

👉 https://maas.xfyun.cn/packageSubscription?inviteCode=MAAS-E4248A94

Order with the invite code, and both parties get equal gift cards. It's like another 50% off.

The offer is limited; discounts won't wait.