B.AI Offers 90% Discount on Top AI Models Including DeepSeek

Overview

B.AI has significantly expanded its discounted model lineup, announcing that five of its most popular AI models are now available at a 90% reduction in cost. Starting September 16 at 17:00 SGT, three new entrants—Qwen3.8-Flash, Hy3, and MiMo-V2.5—officially joined the promotion alongside existing options DeepSeek-V4.1-Flash and GLM-5.3-Flash.

This strategic move positions B.AI as a more accessible gateway to advanced artificial intelligence infrastructure, targeting developers and enterprises looking to optimize their operational budgets. By allowing users to pay just 10% of official pricing, the platform aims to lower the barrier to entry for high-performance reasoning and coding tasks.

With the 90% OFF offer, users pay just 10% of the official pricing for these models, providing significant flexibility to switch between different AI models based on specific workload requirements.

The consolidation of these five models under a single discount umbrella represents a major shift in how cost-efficient AI workflows can be constructed. Industry analysts suggest that such aggressive pricing strategies are becoming increasingly common as competition in the generative AI sector intensifies, forcing providers to offer tangible value beyond mere model availability.

Expanded Model Lineup

The core of B.AI’s latest announcement is the inclusion of three newly discounted models into its existing portfolio. Prior to September 16, the platform already offered deep discounts on DeepSeek-V4.1-Flash and GLM-5.3-Flash. The addition of Qwen3.8-Flash, Hy3, and MiMo-V2.5 completes a diverse suite of tools designed to handle a wide array of computational demands.

DeepSeek-V4.1-Flash remains a standout option for users requiring robust reasoning capabilities. Known for its efficiency and speed, this model is particularly suited for complex problem-solving tasks where latency and accuracy are critical. Its inclusion in the discount program makes it an even more attractive proposition for budget-conscious development teams.

Similarly, GLM-5.3-Flash brings another layer of versatility to the lineup. The GLM series has long been recognized for its strong performance in natural language processing and multilingual tasks. By offering this model at a fraction of its standard cost, B.AI is enabling a broader range of users to leverage its advanced linguistic understanding without incurring prohibitive expenses.

The two newest additions, Qwen3.8-Flash and MiMo-V2.5, further broaden the platform's technical spectrum. Qwen models are frequently praised for their coding abilities and logical reasoning, making them ideal for software development workflows. MiMo-V2.5, while newer to the discounted tier, is expected to offer specialized capabilities that complement the existing models, ensuring that users have options for various niche applications.

Additionally, Hy3 joins the group, rounding out the five-model collection. The presence of such a diverse mix allows users to experiment with different architectures and find the optimal fit for their specific use cases. This diversity is crucial in an era where no single model excels at every task, and the ability to switch seamlessly between them is a significant competitive advantage.

Pricing Structure and Cost Efficiency

The primary driver behind this initiative is the dramatic reduction in cost for end-users. B.AI is structuring its pricing so that customers pay only 10% of the official model rates. This 90% discount applies across the board for the five designated models, creating a uniform and straightforward pricing scheme that is easy for enterprises to forecast and manage.

For many organizations, AI inference costs can escalate rapidly, particularly when dealing with large language models during peak usage periods. By locking in these discounted rates, B.AI is providing a financial safety net that allows teams to innovate without the constant fear of budget overruns. This is especially relevant for startups and small-to-medium enterprises that may have ambitious AI roadmaps but limited capital resources.

The cost efficiency extends beyond simple price cuts. Users gain the freedom to compare different models side-by-side for the same low price point. This encourages a more analytical approach to model selection, where developers can test which model delivers the best results for a particular prompt or dataset before committing to a production workflow.

From coding and reasoning to everyday AI tasks, access to multiple models makes it easier to match your workload with the capabilities you need.

This flexibility is a key component of modern AI infrastructure. Rather than being locked into a single provider or model version, users can dynamically switch between DeepSeek, GLM, Qwen, Hy3, and MiMo depending on the immediate task. For instance, a user might choose DeepSeek for a complex coding challenge, then switch to GLM for drafting technical documentation, all while maintaining a consistent and low cost per token.

Furthermore, this pricing model supports the concept of hybrid workloads. Companies can run lightweight tasks on faster, cheaper models and reserve more powerful (though still discounted) models for heavy lifting. This tiered approach to compute utilization ensures that resources are not wasted on over-specifying solutions for simple problems.

Impact on AI Workflows

The introduction of these discounted models is expected to have a tangible impact on how AI is integrated into daily workflows. Developers are increasingly expected to act as orchestrators of multiple AI agents, selecting the right tool for each sub-task within a larger pipeline. B.AI’s platform facilitates this by making high-quality models economically viable for such granular applications.

In the realm of software engineering, the ability to use DeepSeek-V4.1-Flash and Qwen3.8-Flash at reduced costs can accelerate development cycles. These models are often employed for code generation, debugging, and refactoring. Lowering the cost barrier means that these assistance tools can be used more extensively, potentially leading to higher code quality and faster time-to-market for software products.

Beyond coding, the models support a variety of other critical business functions. Marketing teams might utilize the language capabilities of GLM-5.3-Flash for content creation, while research departments could leverage the reasoning powers of Hy3 for data analysis. The unified platform ensures that these diverse needs can be met without fragmenting the organization across multiple expensive services.

B.AI is positioning itself not just as a model provider, but as an enabler of cost-efficient AI infrastructure. By reducing the financial risk associated with experimenting with new models, the platform empowers users to build more sophisticated and resilient AI systems. This aligns with a broader industry trend toward democratizing access to advanced artificial intelligence technologies.

As the ecosystem continues to evolve, the ability to access top-tier models like DeepSeek at such significant discounts will likely become a standard expectation rather than a novelty. B.AI’s current offering sets a benchmark for value in the market, challenging other providers to consider similar strategies to remain competitive. Ultimately, this benefits the entire community by fostering a more open and economical environment for AI development and deployment.