Unlimited Basic upgrade: Qwen 3.8 27B, GLM and DeepSeek
Published 15 October 2026 by QDivZero Team10 min read
Try QDivZero in action
Create your account
Unlimited Basic gets better for the same €14.95/month: unlimited Qwen 3.8 27B, subject to fair use, plus 5,000 monthly AI Credits for GLM-5.3-Flash, DeepSeek V4.1 Flash and DeepSeek V4 Flash 0731. One subscription for everyday work with your included model and more options for coding, reasoning and agents.
The upgrade matters if you use an AI API regularly. Keep Qwen 3.8 27B as the foundation of your applications and spend premium credits on tasks where you want to explore another capability or get a second opinion. All through QDivZero, without deploying servers or managing GPUs.
Explore Unlimited Basic and its terms · Create a QDivZero account
What the Unlimited Basic upgrade includes
The monthly fee stays the same. The addition is a credit allowance for three premium models, alongside your existing unlimited access to Qwen 3.8 27B.
Unlimited usage applies to Qwen 3.8 27B. Its fair use limits are 60 requests per minute per account/API key and 5 concurrent requests to Qwen 3.8 27B. Premium models draw from one monthly allowance, rather than receiving 5,000 credits each.
What the new AI models bring to your workflow
Coding with an agent can involve understanding a request, reading files, calling tools and reviewing the result. The new models expand your options for these workflows. Below are capabilities and benchmarks published by their developers, followed by our suggestions for using them in Unlimited Basic.
GLM-5.3-Flash: coding with visual context and tools
GLM-5.3-Flash has native multimodal understanding. Z.ai highlights workflows combining code, interface screenshots and visual checks of the result, alongside document and tool use. It is an interesting option to explore for frontend development and tasks involving visual information. Official GLM-5.3-Flash capabilities.
Selected official results:
Sources: Z.ai’s published evaluations and settings and official documentation identifying DeepSWE v1.1. Terminal-Bench 2.1 uses Claude Code 2.1.207; DeepSWE uses mini-swe-agent.
How to use it in the plan: try GLM-5.3-Flash to review an implementation or prepare a change requiring several iterations. It consumes 110 AI Credits per million input tokens and 440 per million output tokens.
DeepSeek V4.1 Flash: reasoning and coding agents
DeepSeek V4.1 Flash combines text and images with improvements aimed at agents handling substantial context. Its technical report describes native support for up to one million tokens, a model capability relevant to analysing extensive material and retaining information throughout tasks with multiple steps. Official DeepSeek V4.1 Flash technical report.
Source: official Instruct model evaluation, using maximum reasoning effort. DeepSeek uses DSH Minimal for Terminal-Bench 2.1 and mini-SWE for DeepSWE v1.1.
In that same table, DeepSeek reports 74.2 on DeepSWE v1.1 against 74.0 for Claude Opus 5.0 and 73.0 for GPT-5.6 Sol. This is a specific coding reference, without implying general equivalence between models. Comparison published by DeepSeek.
How to use it in the plan: reserve it for a difficult bug, an architecture review or an agent task where you want to compare another model’s approach. It consumes 375 AI Credits per million input tokens and 1,500 per million output tokens, so assess the result alongside the credits used.
DeepSeek V4 Flash 0731: another way to allocate your credits
DeepSeek V4 Flash 0731 is the official release replacing the DeepSeek V4 Flash preview. Its release shows specific improvements in terminal tasks, coding and tool use. The comparison with the preview helps explain what this revision brings. Official DeepSeek V4 Flash 0731 release.
Source: DeepSeek’s official table and methodology. Coding agent evaluations use DeepSeek Harness in minimal mode with maximum reasoning effort.
How to use it in the plan: compare it on your regular automation and coding tasks. At 129 AI Credits per million input tokens and 258 per million output tokens, it uses fewer output credits than the other two included premium models for the same number of generated tokens.
How to read these benchmarks
These results are published by each model’s authors. The selected tests cover coding, agents and reasoning, with different evaluation settings. They are not an independent QDivZero evaluation or a comparison under identical conditions. They are also not measurements of our API’s speed.
Native capabilities, modalities and context windows in model cards do not by themselves determine what each endpoint exposes. Check the configuration available in the platform. We recommend testing models on a real task and comparing quality, response time and credit consumption.
Which model should you choose for each task?
Start with this distribution and adjust it based on your project’s results:
For example, develop a function with Qwen 3.8 27B, ask GLM-5.3-Flash to review it and use DeepSeek V4.1 Flash to investigate an unresolved bug. That second opinion is part of the same subscription and draws from the premium allowance.
What AI Credits are and how they are consumed
AI Credits are usage units included in Unlimited Basic. Premium models deduct them according to input and output tokens. Input tokens cover content sent to the model, such as instructions and conversation history; output tokens cover generated content counted in the request.
AI Credits are managed separately from your QDivZero monetary balance, which pays for other models and services at their respective rates. Activate Unlimited Basic without topping up balance, or combine both options. Usage covered by your subscription does not consume that balance.
An example to explain the units: at GLM-5.3-Flash’s rates, a request counting 10,000 input tokens and 2,000 output tokens would consume approximately 1.98 AI Credits: 1.10 for input and 0.88 for output. This is a token-based calculation, rather than a fixed exchange of one credit for one question.
How much can you do with 5,000 AI Credits?
The figures show approximately how many questions with answers, complete conversations or complete tasks you can do with 5,000 AI Credits, depending on the row. Each figure assumes your entire monthly allowance goes to that model and type of use.
For example, ≈900 complete conversations with GLM-5.3-Flash means around 900 chats, each with 10 questions from you and 10 AI replies, if you allocate all 5,000 AI Credits to that use.
Actual usage can vary with question and answer length, conversation context and agent calls.
How your AI Credits are used
Usage is calculated in tokens. Input is what you send to the model, including conversation context; output is what it generates in response.
Total usage adds input and output credits and deducts them from the same monthly allowance of 5,000 AI Credits.
For agent coding, ≈55 with GLM-5.3-Flash represents around 55 complete tasks. Each task can include multiple questions, answers and model calls. Context resent at each step and the work needed to finish affect actual consumption.
How Unlimited Basic compares with other options
The upgrade’s advantage is combining a daily model without per-token billing and an allowance for three premium models under a fixed fee. Build on Qwen 3.8 27B and compare other models without setting up a separate account and billing integration with each provider for that included usage.
Compared with Z.ai and DeepSeek’s metered APIs
Direct APIs are a valid alternative: you pay for what you consume. Here is how that differs from Unlimited Basic:
Official prices checked on 15 October 2026: Z.ai pricing and DeepSeek pricing. Cached inputs have their own rates. These APIs publish prices in US dollars; QDivZero’s subscription is priced in euros. AI Credits are plan usage units and are not converted into money here.
If you only need a few premium model calls, metered usage may cost less. Unlimited Basic is particularly relevant when you use Qwen 3.8 27B regularly and also make use of premium credits. Actual savings depend on your volume and model mix; the monthly fee does not make premium usage unlimited.
Compared with a chat subscription such as Claude Pro
When building an application, check what API access includes. Anthropic distinguishes Claude Pro, focused on the chat experience, from Claude API usage, which is billed separately. Its Max and Team plans do include monthly API credits under their own terms. Official explanation of Claude subscriptions and API billing.
Unlimited Basic includes usage of its models through an API within the plan. Connect your own assistant, application or agent where the tool supports the compatible integration. Your choice depends on whether you want a particular chat experience or to integrate models into your product.
One platform for model integration and request configuration
QDivZero brings these models together with LLM Model Routing and LLM Firewall. Configure routes between destinations and rules for requests from the same platform. These tools support your integration; calls to each model retain their usage terms.
You can also combine the subscription with balance for other QDivZero models or services. Expand your project when needed while keeping Qwen 3.8 27B as your everyday model.
How to activate Unlimited Basic and start using the API
- Create your QDivZero account or sign in to the platform.
- Open Unlimited in the sidebar, choose Unlimited Basic and complete your subscription.
- Check that it is active and copy the identifier for your chosen model from that section.
- Configure your application with a QDivZero API key, the base URL
https://api.qdiv0.com/v1and the subscription model identifier.
Use the identifiers shown in Unlimited for included access. The Qwen 3.8 27B API guide explains the integration with examples; the QDivZero quickstart helps you prepare your first request.
Frequently asked questions about the upgrade
Does Unlimited Basic cost more now?
The fee stays at €14.95/month. It includes unlimited Qwen 3.8 27B subject to fair use and adds 5,000 monthly AI Credits shared across GLM-5.3-Flash, DeepSeek V4.1 Flash and DeepSeek V4 Flash 0731.
Are GLM-5.3-Flash and DeepSeek Flash unlimited too?
All three premium models consume AI Credits. The plan’s unlimited model is Qwen 3.8 27B, subject to fair use. Premium consumption depends on input and output tokens and each model’s credit rate.
What happens if I use all 5,000 AI Credits?
You can keep using unlimited Qwen 3.8 27B, subject to fair use. AI Credits renew monthly. Check your subscription for the exact date, premium access terms and treatment of unused credits. Monetary balance is managed separately.
Do I need a GPU or a model deployment?
Use included models through QDivZero’s API without creating a GPU deployment. You need an active subscription, an API key and the corresponding application configuration.
Which model makes credits last longer?
It depends on how many tokens you send and generate. GLM-5.3-Flash has the lowest input credit rate of the three; DeepSeek V4 Flash 0731 has the lowest output rate. DeepSeek V4.1 Flash consumes more credits per token and may be worthwhile for specific tasks. Compare both the result and the consumption of a complete task.
Start with Qwen 3.8 27B and explore the new models
For €14.95/month, work with unlimited Qwen 3.8 27B subject to fair use and allocate your 5,000 included AI Credits to GLM-5.3-Flash, DeepSeek V4.1 Flash and DeepSeek V4 Flash 0731. Choose a real task, test the models and see which combination works best for your project.
See everything included in Unlimited Basic · Activate Unlimited Basic in QDivZero
