B.AI October Report: Chinese AI Models Lead in Usage and Cost Efficiency

In October, B.AI reached another significant milestone in its platform development, with registered users surpassing 2,920,008 and nearing the 3 million mark. The platform offers more than 70 of the world’s leading multimodal large models, accessible through both API and Web Chat interfaces. As user numbers and model offerings grow in tandem, the platform’s economies of scale continue to become more evident. This growth reflects the growing trust of global developers and users in B.AI’s unified access point and reliable service.

On the model supply side, B.AI has made significant strides this month in multimodal capabilities. ByteDance launched all four models in its Seedance series at once, extending platform capabilities into video creation with support for up to 4K resolution. Meanwhile, new models including GPT-6.1 Terra, Claude Haiku 5.5, MiniMax M3.1 Flash, and Kimi K3.1 Pro have gradually been introduced, further enriching B.AI’s large model ecosystem. On the user experience side, B.AI has ensured that model iterations no longer burden developers: the MiMo-V2.5 series enables seamless migration through automatic routing via existing model IDs, while DeepSeek V4.1 Flash allows users to upgrade to version 1017 without changing model IDs—businesses can immediately access new capabilities without any modifications. Additionally, three rounds of popular model discount offers have been rolled out in succession, with prices as low as 10% off, making high-performance AI access more cost-effective.

The richness of model offerings and enhanced user experiences ultimately manifest in developers' real-world choices. In October, B.AI officially launched its new large model ranking, reflecting the overall rankings and trends of major models on the platform based on actual user usage. The rankings show that Chinese large models demonstrate dominant ecosystem appeal, overwhelmingly occupying the top eight positions on this month’s usage chart, igniting the battle for developer traffic. DeepSeek-V4.1-Flash remains at number one, while MiMo-V2.6-Flash has emerged as the core growth engine this month, showing strong upward momentum.

With the leap in ecosystem scale and continuous evolution of the technological foundation, here are the key advancements by B.AI in October across dimensions such as large model landscape, multimodal infrastructure, and developer benefits:

Competition in the large model arena has entered a critical phase, with developers' actual usage data becoming a key metric for evaluating model performance. In October, the B.AI platform launched its updated large model ranking based on daily volumes of real token consumption and API call frequency. The most significant highlight of this month’s ranking is the comprehensive breakthrough of Chinese large models. They have not only established a clear advantage in cost efficiency but also achieved a leap forward in usage popularity and cutting-edge intelligence, becoming a cornerstone of the global developer ecosystem. From popularity to cost to intelligence, the October ranking reveals the following key trends:

1. At the Peak of Popularity: Chinese large models dominate the top 8 spots on the usage leaderboard, igniting the battle for developer traffic

In the "Call Ranking," which represents absolute influence, Chinese large models have not only secured an overall advantage by occupying the top eight positions, but the competition among the leading models has also become highly intense: the frontrunners continue to solidify their user base, while challengers are rapidly closing the gap through high growth rates, as major vendors engage in a full-scale battle for positioning in high-frequency use cases.

• Absolute leader: DeepSeek-V4.1-Flash consistently ranks #1 on both usage metrics (token volume and request count), reaching a weekly token usage share of 81.1% between October 26 and November 1, establishing an extremely strong traffic moat on the platform.
• Growth engine: Following closely is MiMo-V2.6-Flash from Xiaomi’s ecosystem, with rising usage热度 showing strong upward momentum and challenging the top spot.
• The second tier is closing in rapidly: GLM-5.3-Flash, Qwen3.8-Flash, MiMo-V2.6-Pro, DeepSeek-V4-Flash, MiniMax-M3.1-Flash, and DeepSeek-V4-Pro form a strong chasing group. Leveraging differentiated task performance and highly competitive response capabilities, this second-tier group is quickly capturing traffic in niche scenarios.

2. Diversified cost advantages empower developers to reduce costs across all scenarios

The "Price Ranking" reflects the real-world competition among large models in commercialization. According to October data, international and leading Chinese models have formed a differentiated competitive landscape across various billing dimensions. Meanwhile, cost competition among Chinese large models has entered a deep phase where every fraction of a cent matters, driving the release of extreme cost-performance value.

• International models solidify their lead in the "Comprehensive and Input" benchmarks: GPT-5 Nano tops both the weighted price and input price rankings (from low to high). Meanwhile, Claude Haiku 5.5 and GPT-6 Luna also firmly hold top-five positions in the weighted price ranking and top-four positions in the input price ranking. This demonstrates that international models remain highly competitive as the computational foundation for building large knowledge bases, processing ultra-long documents (RAG), and general intelligent dialogue scenarios.
• Chinese large models dominate the "high-concurrency generation" scenario: In the "output generation" phase, Chinese large models have established a clear structural advantage. MiMo-V2.6-Flash, MiniMax-M3.1-Flash, GLM-5.3-Flash, and Qwen3.8-Flash collectively claim the top four positions with the lowest platform output prices. A direct comparison of the internal pricing of these four leading domestic models reveals the high level of granularity in this competitive segment. The second-ranked MiniMax (USD 0.3 per million tokens) is only USD 0.02 behind the leader, while GLM (USD 0.35 per million tokens) and Qwen (USD 0.376 per million tokens) closely follow with price differences measured in fractions of a cent.

3. Strong breakthrough in intelligent capability: China’s large models have entered the global top tier

Chinese large models do not rely solely on cost-performance advantages; they have also delivered impressive results in the "intelligence ranking" that evaluates logical reasoning and complex task capabilities.

• Strong breakthrough: Kimi K3.1 Pro, introduced by Moonshot, scores 55.1 and firmly enters the top three of the global intelligence ranking, trailing only behind Claude Opus 5.5 and Claude Sonnet 5.5.
• Multiple breakthroughs: Alibaba Cloud's Qwen3.8-Max ranks fifth globally with a score of 53.4. On the forefront of intelligent competition, Chinese large models have firmly secured their place on the global top-tier table.

* For more ranking details and multidimensional data, please visit B.AI Large Model Rankingsto view.

Overall, platform developers have fully moved beyond single-minded price considerations in computing power selection, shifting toward sophisticated, demand-driven scheduling. Whether it’s the “cost-performance engine” supporting high-frequency concurrency or the “high-intelligence brain” powering complex reasoning, large models with different profiles can precisely identify their value niches within B.AI’s real-world workflows.

In October, B.AI made new progress in user base growth, model ecosystem development, and multilingual coverage, further demonstrating platform-scale effects as it steadily transitions from “feature-complete” to “scale-mature.”

• User base continues to grow: registered users reach 2,920,008, surpassing 2.9 million and nearing the 3 million mark;
• The model ecosystem is further enriched: over 70 top-tier large models are centrally integrated through a unified entry point, continuously enhancing model aggregation capabilities;
• Key breakthrough in global capabilities: The B.AI website, Web Chat, and the newly launched Android App v1.3.0 now fully support 11 major international languages. The comprehensive performance upgrades on mobile and the removal of language barriers deliver a smoother, native-level experience in users’ mother tongues and significantly lower the entry barrier for global developers and users.

Behind the growth in scale is a continuously expanding array of models. This month, B.AI continued to enrich its portfolio of high-quality models, with a strong focus on building multi-modal infrastructure. From native visual understanding to joint audio-visual generation, we are committed to enabling developers to seamlessly access top-tier models for text, images, audio, and video—all on a single platform—to precisely meet complex business needs. To help developers more intuitively select the right compute resources, the newly integrated flagship models this month offer comprehensive coverage across all use cases.

1. Seedance series (ByteDance), official group API has been fully integrated

• Model positioning: Flagship multimodal audio-visual joint generation matrix
• Key highlights: Includes four refined models—Seedance 2.5, Seedance 2.0, Seedance 2.0 Fast, and Seedance 2.0 Mini. The flagship 2.5/2.0 versions prioritize high quality and precise control (up to 4K resolution, up to 30-second generation, support for up to 50 assets, and frame-by-frame animation control). The Fast/Mini versions balance generation speed with maximum cost-effectiveness (480p/720p).
• Use cases: Prompt experimentation, rapid iteration of ad storyboards, and short-form video and professional cinematic creation.

2. Kimi K3.1 Pro Series (Moonshot), API and Web Chat both available

• Model positioning: Flagship with ultra-long context and native visual understanding
• Key highlights: Capable of processing up to 1M tokens with powerful native visual parsing. Available in Standard and Highspeed versions (2-3x faster response times).
• Use cases: Complex programming and code refactoring, CAD/3D modeling analysis, lengthy financial research report evaluation, and multi-step Agent workflows with extremely low latency requirements.

3. MiniMax M3.1 Flash (MiniMax), with both API and Web Chat now open

• Model positioning: Powerful tool for complex logical reasoning and large-scale output
• Key highlights: Combines powerful visual reasoning capabilities with support for up to 1M tokens in input context and a maximum output token limit of 512K.
• Use cases: Complex office automation workflows, software engineering and frontend code implementation, long-form presentation writing.

4. Claude Haiku 5.5 (Anthropic), API and Web Chat both available

• Model positioning: Ultra-fast, lightweight text and vision model
• Key highlights: Delivers 1M token context and five adjustable levels of reasoning depth, while maintaining ultra-low latency and exceptional cost-effectiveness—significantly improved over the previous generation in programming, computer operations, and knowledge work.
• Use cases: High-concurrency, low-latency classification, routing, information extraction, sub-agent tasks, and real-time streaming responses.

5. GPT-6.1 Terra (OpenAI), API and Web Chat both available

• Model positioning: High-efficiency visual reasoning and professional text generation model
• Key highlights: Combines adjustable reasoning capabilities with a 1.05M ultra-long context window, supporting up to 128K output tokens, while delivering powerful performance (standard pricing at $1.00/M input, $5.00/M output) and flexible control for complex applications.
• Use cases: Precise generation of computer operation instructions, clear and natural professional writing, and high-frequency visual reasoning workflows.

* LoginB.AI, unlock top-tier multimodal models and start your intelligent experience today.

As the number of models continues to grow and iteration speeds accelerate, the core challenge for developers has shifted from “insufficient available models” to “excessively high adoption costs.” In response, B.AI is simultaneously enhancing efforts across integration, migration, and upgrade processes to continuously reduce the burden of model updates on businesses.

1. Seamless Migration: MiMo-V2.5 Automatic Routing Ensures Business Continuity

In response to the transition from the MiMo-V2.5 series to the next-generation model, B.AI has introduced a seamless migration solution: the platform leverages its underlying scheduling capabilities to automatically route requests from legacy model IDs while simultaneously aligning with the new billing standards. Additionally, developers are provided with ample time to adapt and test their code, ensuring uninterrupted service for all applications until the legacy IDs are officially retired.

2. ID-free upgrade: Upgrade DeepSeek V4.1 Flash to version 1017

The upgrade to DeepSeek V4.1 Flash works the same way: after updating to the latest version 1017, users can immediately experience enhanced capabilities without changing their model ID—performance improves across all 11 benchmarks, with the highest pass rate for DeepSWE v1.1 reaching 77.0%, faster responses, more concise outputs, shorter processing times for similar tasks, and optimized agent and programming capabilities, all while maintaining existing integration methods for stronger performance and higher efficiency.

3. B.AI Web Chat adds model search functionality

Faced with the decision fatigue from 70+ models, B.AI Web Chat has added a model search feature—simply type the name in the model selection window to instantly locate your desired model, transitioning from overwhelming choices to precise targeting for faster switching and easier use.

In October, B.AI launched multiple rounds of promotions around popular large models, further lowering the barrier and cost for developers to access high-performance models.

1. First-round multi-tier discount matrix: Starting October 4, 10:00 (SGT)

The multi-tier discount matrix for popular models is now live: MiMo-V2.6-Flash is available at 10% off, leveraging its existing multimodal and high-throughput architecture to become the preferred choice for high-concurrency encoding and Agent workflows; Qwen3.8-Flash at 30% off, MiMo-V2.6-Pro at 50% off, DeepSeek-V4.1-Flash at 60% off, and GLM-5.3-Flash at 70% off—all effective immediately.

2. Second Round of Promo Matrix: Starting October 22, 17:00 (SGT)

A new promotional matrix is here: DeepSeek-V4.1-Flash-1017 is now 40% off, with全面提升 benchmark performance and significantly enhanced programming and Agent capabilities—more value without higher pricing; Qwen3.8-Flash is 20% off, and GLM-5.3-Flash is 30% off, covering long-text, ultra-fast inference, and complex multimodal needs.

3. Third promotion: MiniMax-M3.1-Flash at 50% off for a limited time

The newly popular model MiniMax-M3.1-Flash (Space Bunny) is now offering a limited-time discount: during the promotion period, enjoy 50% off on the B.AI web platform, Android app, and API, helping developers complete complex tasks—from code development to multi-step automated workflows—at a lower cost.

Leveraging its scalable computing throughput and unified API access, B.AI will continuously integrate high-quality global model resources, offering developers more competitive, accessible pricing and a richer computing matrix to drive the low-cost commercialization of AI technology.

*Follow B.AI Official X to get the latest model updates and exclusive compute power deals第一时间

Join B.AI Official Community to unlock more exclusive benefits and technical support

Looking back at October, B.AI once again demonstrated the pivotal value of unified foundational infrastructure in the AI era, with a user base nearing 3 million and an ecosystem of over 70 multimodal large models. From the release of the large model rankings to comprehensive expansion of multimodal computing power, and from seamless, frictionless integration to ongoing democratization of computing resources, B.AI has consistently focused on abstracting technical complexity and breaking down cost barriers to computing. Looking ahead, B.AI will continue to deepen its global infrastructure development, leveraging a unified API entry point and massive throughput advantages to collaborate with leading global large model providers and developers alike, jointly accelerating the full-scale explosion of the next generation of super-applications and the Agent economy.