Bean Pack announced access to multiple third-party models and simultaneously introduced time-limited discounts, lowering the frequency of some models.Author of the book
Editor's note:
It's also a free-for-all model.
On November 12, Bean Package announced access to multiple third-party models such as Kimi K3.1 Pro, GLM 5.3, and DeepSeek V4.1 Flash to support user-on-demand completion of tasks.
At the same time, the platform offers two time limits: one for the use of the activity model during the designated times of weekdays and holidays; the other for the beans pack 2.1 Lite, DeepSeek V4.1 Flash.
01
Completing multiple models and solving "there are options"
In this round of updates, the beanbag works with one-time access to Kimi 3.1 Pro, GLM 5.3, DeepSeek V4.1 Flash and three third-party models, further complementing the matrix of model capabilities.
Specifically, the three models are focused on different business-level work scenarios: Kimi 3.1 Pro mainly focuses on the context processing capability of over a million words; GLM 5.3 focuses on Chinese language understanding and generation; DeepSeek V4.1 Flash focuses on lightweight everyday tasks with high response speeds and high cost-effectiveness.
In addition to the third-party models that have already been opened, and the mainstream models such as the byte-bouncing self-developed beanbag 2.1 and beanbag 2.1 Lite, beanbag work already has a three-tiered model matrix covering "high-end capability, balanced general-purpose, and light-efficiency".
Different capabilities, different pricing models complement each other, basically covering the needs of multi-scenario AI for enterprises from core complex tasks to everyday basic work.
In fact, soybean work has already opened up third-party modeling capabilities to enterprise customers. This round of expansion is not a mere quantity superimposition, but a full-scale scenario replacement based on real business workflows.
Image source: The official website of Bean Bags
In addition, we also noticed that soybean packaging work is simultaneously updating the LOGO, and the familiar image of soybean packaging for this "certificate of increase" has entered the professional office scene.
02
Transparency in the quotas,
Putting cost decision-making back to users
If the model expansion is to solve the problem of "whether or not to choose", then the transparency marking of the rate of incremental consumption is to solve the problem of "whether or not to choose clearly".
In today's enterprise-level AI marketplace, cost opacity is a common problem for many businesses. The billing rules for different models vary, with some billing based on input and output tokens, some billing based on the number of calls, and switching between different units is complex, making it difficult for ordinary users to quickly judge the actual cost of using a model.
Many teams often find it difficult to predict the cost of a single task in the process of use, and it is difficult to trace the specific composition of the cost beyond the budget until the bill comes out.
To address this problem, the bean bag work is unified on the model selection page, marking the lower the share consumption ratio, the higher the coefficient, and the lower the platforms that consume the same tasks. Under the consumability standard of all models to the same quantity system, user selection models can intuitively compare cost differences, combine task priorities with budgetary quick decisions, and facilitate the cost management of enterprises as departments.
On the basis of simple, small-scale tasks free of charge, soybean packages work simultaneously launched two limited-time discount policies, further reducing the cost of use:
One is the use of the activity model during the specified time discount, weekdays and holidays, which can benefit from a reduction in the amount of consumption, taking into account the user cost and the computational efficiency of the platform; the second is the reduction in the frequency of the high frequency model, the lightweight model for the daily high frequency use of the beanbag 2.1 Lite and the DeepSeek V4.1 Flash, the limit reduction in the frequency of consumption, the consumption of the same tasks less, improving the daily use of all users.
As the basic capabilities of large models gradually flatten, how to help enterprises use AI to budget is the key to landing. From providing a single model, to providing multiple model matrices; from black box billing, to transparency cost control, the product transformation path of bean bag work, fits the transformation trend of enterprise AI from "taste fresh" to "standard production tools".
Looking ahead, when the model is large enough, the frequency is clear enough, the user may not have to worry about "which model to choose", and the platform based on task-type automatic routing, intelligent scheduling, will be the next natural product direction.
* Source of the image: Beans Bags work website
A question from the PolesHow do you see the product transformation path of the soybean bag?
