B.AI October Ecological Monthly Report | Multimodal flagship matrix fully expanded, new entries on the large model leaderboard witness the comprehensive explosion of China's large models

In October, B.AI welcomed another important milestone in platform development. The number of registered users surpassed 2,920,008, nearing the 3 million mark. The platform brings together over 70 top global multimodal models, open to the public through API and Web Chat dual access. The user base and model supply are rising in tandem, with scale effects continuously manifesting on the platform. Behind this growth is the choice made by global developers and users for B.AI's unified entry and stable services.

In terms of model supply, B.AI made significant strides in multimodal capabilities this month. ByteDance's Seedance series launched four video generation models at once, extending the platform's capabilities to video creation, with a maximum output of 4K quality. Meanwhile, new products such as GPT-6.1 Terra, Claude Haiku 5.5, MiniMax M3.1 Flash, and Kimi K3.1 Pro have successively landed, further enriching the B.AI large model matrix. On the user experience front, B.AI has made model iteration no longer a burden for developers. The MiMo-V2.5 series achieves smooth migration through automatic routing with old model IDs, and DeepSeek V4.1 Flash can upgrade to version 1017 without changing model IDs, allowing businesses to access new capabilities without modifications. At the same time, three rounds of popular model discount rights have been launched, with prices as low as 10%, making high-performance AI calls more cost-effective.

The richness of model supply and the upgrade of user experience ultimately reflect the real choices of developers. In October, B.AI's new version of the large model ranking officially went live, presenting the overall ranking and trend changes of major models on the platform based on real user call data. The rankings show that Chinese large models exhibit a dominating ecological attractiveness, strongly occupying the top eight of this month's call rankings, igniting a competition for developer traffic. Among them, DeepSeek-V4.1-Flash firmly holds the top position; MiMo-V2.6-Flash, with its strong upward momentum, has become the core growth engine of the month.

With the leap in ecological scale and the continuous evolution of the technical foundation, here are the key developments of B.AI in October across dimensions such as large model landscape, multimodal infrastructure, and developer rights:

1. New Large Model Ranking: Seeing the Model Landscape through Real Calls

The competition in the large model track has entered deep waters, with real call data from developers becoming an important standard for assessing comprehensive model strength. In October, the B.AI platform released a new large model ranking based on daily massive real token consumption and API call frequency. The most significant highlight of this month's ranking is undoubtedly the comprehensive explosion of Chinese large models. They have not only established significant advantages in cost control but have also achieved leapfrog overtaking in call popularity and cutting-edge intelligence levels, becoming the backbone of the global developer ecosystem.

From popularity, cost to intelligence, the October ranking presents the following core trends:

1. Pinnacle of Popularity: Chinese Large Models Sweep the Top 8 of Call Rankings, Igniting Developer Traffic Competition

In the "call ranking," which represents absolute influence, Chinese large models have not only established an overall advantage by sweeping the top eight, but the ranking competition among leading models also shows a highly competitive situation: leaders continue to consolidate their traffic base, while followers are quickly closing in with high growth rates, and a positional battle around high-frequency business scenarios has fully unfolded among major manufacturers.

• Absolute Dominance: DeepSeek-V4.1-Flash firmly holds the top position in both call rankings (token usage and request counts). From October 26 to November 1, its weekly token usage share reached 81.1%, establishing a deep traffic moat within the platform.

• Growth Engine: Following closely is MiMo-V2.6-Flash from the Xiaomi ecosystem, which shows a strong upward trend in call popularity, challenging the top position.

• Second Tier in Close Pursuit: Following closely are GLM-5.3-Flash, Qwen3.8-Flash, MiMo-V2.6-Pro, DeepSeek-V4-Flash, MiniMax-M3.1-Flash, and DeepSeek-V4-Pro, forming a strong chasing tier. Relying on differentiated task performance and highly competitive response performance, the second-tier camp quickly captures traffic in segmented scenarios.

2. Diversified Cost Advantages Empower Developers to Reduce Costs Across All Scenarios

The "price ranking" reflects the real competition of global large models in commercial implementation. From the data in October, a dislocated competitive pattern has formed between international and Chinese leading models across different billing dimensions. Meanwhile, internal cost competition among Chinese large models has entered a "penny-pinching" deep water zone, driving the release of extreme cost-effectiveness.

• International Models Solidify "Comprehensive and Input" Benchmarks: GPT-5 nano strongly tops both the weighted price and input price rankings (from low to high). At the same time, Claude Haiku 5.5 and GPT-6 Luna also remain in the top five of the weighted price ranking and the top four of the input price ranking. This indicates that in constructing massive knowledge bases, conducting ultra-long document reading (RAG), and general intelligent dialogue scenarios, international models remain a highly competitive computational foundation.

• Chinese Large Models Dominate "High Concurrency Generation" Scenarios: In the "output generation" phase, Chinese large models have established significant structural advantages. MiMo-V2.6-Flash, MiniMax-M3.1-Flash, GLM-5.3-Flash, and Qwen3.8-Flash strongly occupy the top four positions for the lowest output prices on the platform. Comparing the internal pricing of these four leading domestic models, one can directly observe the refinement of competition in this range. The second-ranked MiniMax (0.3 USD/million tokens) has only a slight price difference of 0.02 USD from the top, while GLM (0.35 USD/million tokens) and Qwen (0.376 USD/million tokens) closely follow with cent-level price differences.

3. Strong Breakthrough in Intelligence Levels: Chinese Large Models Enter the Global First Tier

Chinese large models do not only win through cost-effectiveness; they also deliver impressive results in the "intelligence ranking," which assesses logical reasoning and complex task capabilities.

• Strong Breakthrough: Kimi K3.1 Pro, launched by Moonshot, with a score of 55.1, strongly enters the top three of the global intelligence ranking (only behind Claude Opus 5.5 and Claude Sonnet 5.5).

• Blooming in Multiple Areas: Alibaba Cloud's Qwen3.8-Max, with a score of 53.4, firmly holds the fifth position globally. In the cutting-edge intelligence arena, Chinese large models have firmly established themselves at the global first-tier table.

For more ranking details and multidimensional ranking data, please visit B.AI Large Model Ranking.

Overall, platform developers have completely transcended single price considerations in computational power selection, shifting towards refined on-demand scheduling. Whether supporting high-frequency concurrency as a "cost-effective engine" or driving complex reasoning as a "high-intelligence brain," different positioned large models can precisely lock in their value ecological niche within B.AI's real business flow.

2. Core Data and Platform Milestones: Scale Effects Continue to Amplify

In October, B.AI achieved new progress in user scale, model ecology, and multilingual coverage, with the platform's scale effects further manifesting, steadily moving from "fully functional" to "scale maturity."

• User Scale Continues to Expand: Registered users reached 2,920,008, surpassing 2.9 million and nearing the 3 million mark;

• Model Ecology Further Enriched: Over 70 top large models have been integrated into a unified entry, with model aggregation capabilities continuously enhanced;

• Globalization Capabilities Achieve Key Breakthroughs: The B.AI official website, Web Chat, and the newly launched Android App v1.3.0 fully support 11 major international languages. The comprehensive upgrade of mobile performance and the breaking of language barriers not only provide a smoother native-level experience but also significantly lower the access threshold for global developers and users.

3. Expansion of Top Model Matrix: Strengthening Multimodal Infrastructure to Build All-Scenario Computational Supply

Behind the scale growth is a continuously enriched model supply. This month, B.AI continued to expand its quality model matrix, focusing on building "multimodal" infrastructure. From native visual understanding to audio-visual joint generation, we are committed to enabling developers to seamlessly call top models that handle text, images, and audio-video across all dimensions on one platform, precisely fitting complex business needs. To allow developers to more intuitively select suitable computational power, the newly integrated heavyweight models this month achieve full-scenario coverage in positioning and capabilities.

1. Seedance Series (ByteDance), Official Group API Fully Integrated

• Model Positioning: Flagship matrix for multimodal audio-visual joint generation

• Core Highlights: Includes four refined models: Seedance 2.5, Seedance 2.0, Seedance 2.0 Fast, and Seedance 2.0 Mini. The 2.5/2.0 flagship versions focus on high quality and strong control (up to 4K resolution, maximum 30 seconds generation, supporting up to 50 materials and first and last frame animation control); Fast/Mini versions balance generation speed and extreme cost-effectiveness (480p/720p).

• Applicable Scenarios: Prompt word trial and error, rapid iteration of advertising storyboards, short video and professional film-level creation.

2. Kimi K3.1 Pro Series (Moonshot), API and Web Chat Dual Access

• Model Positioning: Flagship for ultra-long context and native visual understanding

• Core Highlights: Equipped with 1M token ultra-long processing capability and powerful native visual analysis. Offers standard and Highspeed versions (response speed improved by 2-3 times).

• Applicable Scenarios: Complex programming and code restructuring, CAD/3D modeling analysis, long financial report analysis, and multi-step agent workflows that are extremely sensitive to delays.

3. MiniMax M3.1 Flash (MiniMax), API and Web Chat Dual Access

• Model Positioning: A powerful tool for complex logical reasoning and large output

• Core Highlights: Combines strong visual reasoning capabilities, supports 1M ultra-long context input while providing an output token limit of up to 512K.

• Applicable Scenarios: Complex office automation execution, software engineering and frontend code implementation, long-form content creation for presentations.

4. Claude Haiku 5.5 (Anthropic), API and Web Chat Dual Access

• Model Positioning: Ultra-fast lightweight text and visual model

• Core Highlights: Provides 1M token context and five adjustable levels of thinking depth while maintaining extremely low latency and high cost-effectiveness, showing significant improvements in programming, computer operations, and knowledge work compared to the previous generation.

• Applicable Scenarios: High concurrency, latency-sensitive classification, routing, information extraction, and sub-agent tasks, as well as real-time streaming responses.

5. GPT-6.1 Terra (OpenAI), API and Web Chat Dual Access

• Model Positioning: Efficient visual reasoning and professional text creation model

• Core Highlights: Combines adjustable reasoning capabilities with a 1.05M ultra-long context window, supports up to 128K output tokens, while providing strong performance (standard pricing $1.00/M input, $5.00/M output) and flexible control over complex applications.

• Applicable Scenarios: Precise computer operation instruction generation, clear and natural professional text writing, and daily high-frequency visual reasoning workflows.

* Log in to B.AI, unlock top-tier models across dimensions, and start your multimodal intelligent experience immediately.

4. Technical Experience Leap: Smooth Access, Seamless Switching

As the number of models continues to increase and the iteration pace accelerates, the core challenge faced by developers has shifted from "insufficient optional models" to "high follow-up costs." In response, B.AI has focused on access, migration, and upgrade processes to continuously reduce the burden of model updates on business.

1. Smooth Migration: MiMo-V2.5 series automatic routing ensures business continuity

In the face of the migration from the MiMo-V2.5 series to the new generation of models, B.AI has introduced a smooth transition plan: the platform leverages underlying scheduling capabilities to achieve automatic routing for old model IDs while synchronously matching the new billing standards. At the same time, it reserves ample code adaptation and testing windows for developers, ensuring uninterrupted service for various applications before the old ID is officially discontinued.

2. ID-Free Upgrade: DeepSeek V4.1 Flash upgraded to version 1017

The upgrade of DeepSeek V4.1 Flash is similar; after upgrading to the latest version 1017, users can directly experience new capabilities without changing the model ID: 11 benchmark tests show comprehensive improvements, with DeepSWE v1.1 achieving a maximum pass rate of 77.0%, faster responses, more refined outputs, shorter task durations, and optimized agent and programming capabilities, allowing users to gain stronger performance and higher efficiency while maintaining their existing access methods.

3. B.AI Web Chat adds model search function

In response to the selection costs brought by over 70 models, B.AI Web Chat has added a model search function, allowing users to quickly locate target models by entering names in the model selection window, making switching more efficient and usage more convenient.

5. Inclusive Computing Power and Developer Rights: Making High-Performance AI Access More Cost-Effective

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

1. First Round of Multi-Tier Discounts: Starting October 4 at 10:00 (SGT)

The multi-tier discount matrix for popular models is launched: MiMo-V2.6-Flash enjoys a 90% discount, leveraging its existing multimodal and high-throughput architecture, becoming the preferred choice for high-concurrency coding and agent workflows; Qwen3.8-Flash enjoys a 70% discount, MiMo-V2.6-Pro enjoys a 50% discount, DeepSeek-V4.1-Flash enjoys a 40% discount, and GLM-5.3-Flash enjoys a 30% discount, all effective simultaneously.

2. Second Round of Discount Matrix: Starting October 22 at 17:00 (SGT)

The new round of discount matrix takes over: the upgraded DeepSeek-V4.1-Flash-1017 enjoys a 40% discount, with benchmark test performance comprehensively improved, programming and agent capabilities significantly enhanced, achieving "more for less"; Qwen3.8-Flash enjoys a 20% discount, GLM-5.3-Flash enjoys a 30% discount, covering long texts, ultra-fast reasoning, and complex multimodal needs.

3. Third Round of Discounts: MiniMax-M3.1-Flash limited-time 50% discount

The newly popular model MiniMax-M3.1-Flash (Space Bunny) welcomes a limited-time benefit: during the event, B.AI web, Android App, and API all enjoy a 50% discount, helping developers complete complex tasks from code development to multi-step automation workflows at a lower cost.

Leveraging the advantages of large-scale computing power throughput and unified API access, B.AI will continue to integrate high-quality model resources globally, providing developers with more competitive inclusive pricing and a richer computing power matrix, continuously promoting the low-cost commercialization of AI technology.

* Follow B.AI Official X to unlock cutting-edge model dynamics and exclusive computing power discounts in real-time.

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

Looking back at October, B.AI, with nearly 3 million users and an ecosystem matrix of over 70 multimodal large models, once again confirms the hub value of unified underlying infrastructure in the AI era. From the release of the large model leaderboard to the comprehensive expansion of multimodal computing power, to the ultra-simple seamless access experience and continuously released inclusive computing power, B.AI is committed to folding technical complexity and breaking down the barriers of computing power costs. Looking to the future, B.AI will continue to deepen global infrastructure construction, relying on unified API access and large-scale throughput advantages, partnering with top global large model vendors and a wide range of developers to accelerate the comprehensive explosion of the next generation of super applications and the agent economy.