B.AI October Ecosystem Monthly Report | Full Expansion of the Multimodal Flagship Matrix, New Additions to the Large Model Leaderboard Witness the Rapid Rise of Chinese Large Models

In October, B.AI reached another important milestone in its platform development. The number of registered users surpassed 2,920,008, nearing the 3 million mark. The platform aggregates over 70 top global multimodal foundational models, opening access to the public through dual channels: API and Web Chat. As user scale and model supply rise synchronously, economies of scale continue to manifest on the platform. Behind this growth lies the choice made by global developers and users toward B.AI’s unified entry point and stable service.

Regarding model supply, B.AI advanced deeply into multimodal capabilities this month. ByteDance's Seedance series launched four video generation models simultaneously, extending platform capabilities to video creation with outputs up to 4K resolution. Meanwhile, new products such as GPT-6.1 Terra, Claude Haiku 5.5, MiniMax M3.1 Flash, and Kimi K3.1 Pro were gradually added, further enriching B.AI's foundational model matrix. On the experience front, B.AI ensures that model iterations no longer burden developers. The MiMo-V2.5 series enables smooth migration through automatic routing based on legacy model IDs, while DeepSeek V4.1 Flash can be upgraded to version 1017 without changing the model ID, allowing businesses to adopt new capabilities without modifying existing code. Concurrently, three rounds of promotional discounts for popular models were introduced sequentially, with prices dropping to as low as 10% of the original rate, making high-performance AI invocations significantly more cost-effective.

The enrichment of model supply and the upgrade in user experience are ultimately reflected in the real-world choices of developers. In October, B.AI officially launched its newly revamped foundational model ranking list, which bases its metrics on actual user invocation data to display overall rankings and trend shifts across models. The list reveals that Chinese foundational models demonstrate dominating ecological appeal, sweeping the top eight spots on this month's invocation rankings and igniting a fierce battle for developer traffic. Among them, DeepSeek-V4.1-Flash firmly holds the top position, while MiMo-V2.6-Flash has emerged as the core engine for rapid growth this month due to its strong upward momentum.

Alongside the leap in ecological scale and the continuous evolution of the technical foundation, here are B.AI's key advancements in October across dimensions such as foundational model landscapes, multimodal infrastructure, and developer benefits:

I. New Foundational Model Rankings: Observing the Model Landscape Through Real Invocations

Competition in the foundational model track has entered deep waters, where developers' actual invocation data has become a crucial benchmark for evaluating comprehensive model capabilities. In October, the B.AI platform, based on massive daily real-world token consumption and API invocation frequency, heavily released the newly updated foundational model rankings. The most core highlight of this month's list undoubtedly is the full-scale eruption of Chinese foundational models. They have not only established significant advantages in cost control but also achieved leapfrogged advancements in invocation popularity and cutting-edge intelligence levels, becoming the backbone of the global developer ecosystem. From popularity and cost to intelligence, the October rankings reflect the following core trends:

1. Peak Popularity: Chinese Foundational Models Sweep Top 8 on Invocation Rankings, Igniting the Battle for Developer Traffic

In the "Invocation Rankings," which represent absolute influence, Chinese foundational models have not only secured an overarching advantage by sweeping the top eight, but the competition among leading models also shows highly intense positioning: frontrunners continuously consolidate their traffic base, while chasers rapidly close the gap with high growth rates. The positioning battle among major vendors around high-frequency business scenarios has been fully unleashed.

Absolute Dominator: DeepSeek-V4.1-Flash firmly holds the top spot on both sides of the invocation rankings (token volume and request count). From October 26 to November 1, its weekly token usage share reached 81.1%, establishing an exceptionally wide traffic moat within the platform.

Growth Engine: Trailing closely behind is Xiaomi ecosystem's MiMo-V2.6-Flash, whose invocation popularity shows strong upward momentum and is currently mounting a challenge to the top spot.

Dense Second Tier Chasing: The following tightly packed group—GLM-5.3-Flash, Qwen3.8-Flash, MiMo-V2.6-Pro, DeepSeek-V4-Flash, MiniMax-M3.1-Flash, and DeepSeek-V4-Pro—forms a powerful chasing tier. Relying on differentiated task performance and highly competitive response times, the second-tier factions quickly capture traffic from niche scenarios.

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

The "Price Rankings" reflect the real-world commercialization race among global foundational models. Data from October indicates that international and top Chinese models have formed an asymmetric competitive landscape across different billing dimensions. Meanwhile, internal cost competition among Chinese foundational models has entered a highly granular "battle over fractions of a cent" phase, driving the release of ultimate cost-effectiveness.

International Models Solidify "Composite & Input" Benchmarks: GPT-5 nano forcefully tops both the weighted price and input price rankings (ordered low to high). Meanwhile, Claude Haiku 5.5 and GPT-6 Luna also firmly hold top five positions in weighted pricing and top four in input pricing. This demonstrates that in scenarios like building massive knowledge bases, processing ultra-long documents (RAG), and general intelligent dialogue, international models remain highly competitive computational backbones.

Chinese Foundational Models Dominate "High-Concurrency Generation" Scenarios: In the "output generation" segment, Chinese models have established a pronounced structural advantage. MiMo-V2.6-Flash, MiniMax-M3.1-Flash, GLM-5.3-Flash, and Qwen3.8-Flash strongly sweep the bottom four spots for platform output pricing. Comparing the internal pricing of these four leading domestic models visually reveals the granularity of competition in this range. Runner-up MiniMax ($0.30/million tokens) trails the leader by a mere $0.02, while GLM ($0.35/million tokens) and Qwen ($0.376/million tokens) relentlessly pursue at penny-level margins.

3. Powerful Breakthrough in Intelligence: Chinese Foundational Models Join the Global First Tier

Chinese foundational models do not win solely on cost-effectiveness; they also delivered outstanding results in the "Intelligence Rankings," which evaluate logical reasoning and complex task capabilities.

Powerful Breakthrough: Kimi K3.1 Pro, launched by Moonshot, aggressively broke into the global intelligence top three with a high score of 55.1 (ranking immediately after Claude Opus 5.5 and Claude Sonnet 5.5).

Blooming Across Multiple Fronts: Alibaba Cloud's Qwen3.8-Max steadily secures fifth place globally with a score of 53.4. At the most frontier intelligence arena, Chinese foundational models have firmly taken their seat at the table of the global first tier.

* For more ranking details and multidimensional leaderboard data, please visit theB.AI Foundational Model Leaderboard.

Overall, platform developers have completely moved past singular price considerations in computational resource selection, shifting toward fine-grained, on-demand scheduling. Whether it is the "cost-effective engine" supporting high-frequency concurrency or the "high-intelligence brain" driving complex reasoning, differently positioned foundational models can precisely lock into their own value ecosystems within B.AI's real-world business workflows.

II. Core Data & Platform Milestones: Economies of Scale Continue to Amplify

In October, B.AI achieved new progress across dimensions such as user scale, model ecology, and multilingual coverage. Platform economies of scale are further materializing, steadily advancing from "feature-complete" toward "scale maturity."

Continuously Expanding User Base: Registered users reached 2,920,008, breaking through 2.9 million and approaching the 3 million milestone;

Further Enriched Model Ecology: Over 70 top foundational models are centrally integrated into a unified entry point, continuously strengthening model aggregation capabilities;

Key Breakthrough in Globalization Capabilities: B.AI's official website, Web Chat, and the newly launched Android App v1.3.0 fully support 11 mainstream international languages. Comprehensive performance upgrades on mobile devices and the breaking of language boundaries not only deliver a smoother native-language-level experience but also comprehensively lower the entry threshold for global developers and users.

III. Expansion of Top-Tier Model Matrix: Focusing on Multimodal Infrastructure to Build Full-Scenario Computational Supply

Behind scalable growth lies continuously enriched model supply. This month, B.AI continues to expand its high-quality model matrix, with a strategic focus on "multimodal" infrastructure construction. From native visual understanding to audio-visual joint generation, we are committed to enabling developers to seamlessly invoke full-dimensional top-tier models for handling text, images, and audio/video within a single platform, precisely aligning with complex business requirements. To help developers more intuitively select appropriate computational power, the heavyweight models onboarded this month achieve full-scenario coverage in both positioning and capabilities.

1. Seedance Series (ByteDance): Official Group APIs Now Fully Integrated

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

Core Highlights: Features 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 emphasize high image quality and strong control (up to 4K resolution, max 30-second generation, supporting up to 50 assets and start/end frame animation control); the Fast/Mini versions balance generation speed with ultimate cost-effectiveness (480p/720p).

Applicable Scenarios: Prompt experimentation, rapid iteration of advertising storyboards, short-form video production, and professional cinematic-grade creation.

2. Kimi K3.1 Pro Series (Moonshot): Open on Both API and Web Chat

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

Core Highlights: Features 1M token ultra-long processing capacity and powerful native visual parsing. Available in Standard and Highspeed versions (response speed increased by 2-3x).

Applicable Scenarios: Complex programming and code refactoring, CAD/3D modeling parsing, lengthy financial research report analysis, and highly latency-sensitive multi-step Agent workflows.

3. MiniMax M3.1 Flash (MiniMax): Open on Both API and Web Chat

Model Positioning: Tool for complex logical reasoning and ultra-large output

Core Highlights: Combines powerful visual reasoning capabilities, supports 1M ultra-long context input, and provides an ultra-large output token limit of up to 512K.

Applicable Scenarios: Complex office automation execution, software engineering and frontend code implementation, and long-form presentation drafting.

4. Claude Haiku 5.5 (Anthropic): Open on Both API and Web Chat

Model Positioning: Ultra-fast lightweight text and vision model

Core Highlights: Maintains extremely low latency and high cost-effectiveness while offering 1M token context and five adjustable depth-of-thought levels, marking a significant improvement over the previous generation in programming, computer operations, and knowledge work.

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

5. GPT-6.1 Terra (OpenAI): Open on Both API and Web Chat

Model Positioning: Efficient visual reasoning and professional text creation model

Core Highlights: Combines adjustable reasoning capabilities with a 1.05M ultra-long context window, supporting up to 128K output tokens. While delivering robust performance (standard pricing: $1.00/M input, $5.00/M output), it offers 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.

* VisitB.AI to unlock full-dimensional top-tier foundational models and immediately begin your multimodal intelligent experience.

IV. Technical Experience Leap: Smooth Integration, Seamless Switching

As the number of models continues to increase and iteration pace accelerates, the core challenge faced by developers has shifted from "insufficient selectable models" to "excessively high follow-up costs." In response, B.AI synchronizes efforts across integration, migration, and upgrade processes, continuously reducing the burden that model updates impose on business operations.

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

Facing the migration of the MiMo-V2.5 series to next-generation models, B.AI has introduced a smooth transition solution: leveraging underlying scheduling capabilities, the platform automatically routes legacy model IDs while synchronously applying new billing standards. Adequate windows for code adaptation and testing are reserved for developers, ensuring uninterrupted application services until the legacy IDs are officially decommissioned.

2. ID-Free Upgrade: DeepSeek V4.1 Flash Upgraded to Version 1017

The upgrade for DeepSeek V4.1 Flash operates similarly; upon upgrading to the latest version 1017, users can directly experience new capabilities without changing the model ID. Performance across 11 benchmark tests is comprehensively elevated, with the highest tier of DeepSWE v1.1 achieving a 77.0% pass rate. Responses are faster and outputs more concise, reducing time consumption for similar tasks, while Agent and programming capabilities are also optimized. Users gain stronger performance and higher efficiency while maintaining their current integration methods.

3. B.AI Web Chat Adds Model Search Functionality

Confronting the selection overhead brought by 70+ models, B.AI Web Chat now features a model search function. Simply entering a name in the model selection window quickly locates the target model. Moving from massive options to precise direct access makes switching more efficient and usage more convenient.

V. Inclusive Computing Power & Developer Benefits: Making High-Performance AI Invocations More Cost-Effective

In October, B.AI consecutively rolled out multiple rounds of promotions targeting popular foundational models, further lowering the threshold and cost for developers invoking high-performance models.

1. First Round Multi-Tier Discount Matrix: Effective from 10:00 SGT, October 4

The multi-tier discount matrix for popular models launched first: MiMo-V2.6-Flash priced at 10% of the original rate leverages its original full-modality and high-throughput architecture to become the preferred choice for high-concurrency coding and Agent workflows; Qwen3.8-Flash priced at 30%, MiMo-V2.6-Pro priced at 50%, DeepSeek-V4.1-Flash priced at 60%, and GLM-5.3-Flash priced at 70% all take effect simultaneously.

2. Second Round Discount Matrix: Effective from 17:00 SGT, October 22

A new round of discounts takes the baton: the upgraded DeepSeek-V4.1-Flash-1017 priced at 60% of the original rate delivers comprehensively elevated benchmark performance, significantly enhanced programming and Agent capabilities, achieving "more capability at the same price"; Qwen3.8-Flash priced at 80% and GLM-5.3-Flash priced at 70% cover long-text processing, ultra-fast reasoning, and complex multimodal needs.

3. Third Round Promotion: Limited-Time 50% Pricing for MiniMax-M3.1-Flash

The newly emerging popular model MiniMax-M3.1-Flash (Space Bunny) introduces a limited-time benefit: throughout the campaign period, B.AI's web portal, Android App, and API endpoints all offer pricing at 50% of the original rate, empowering developers to complete complex tasks ranging from code development to multi-step automation workflows at a lower cost.

Leveraging scaled computational throughput advantages and unified API integration, B.AI will continuously aggregate global premium model resources, providing developers with more competitive inclusive pricing and a richer computational matrix, persistently driving the low-cost commercial adoption of AI technologies.

* FollowB.AI's Official X Account to unlock cutting-edge model updates and exclusive computing power deals promptly.

* JoinB.AI's Official Community to unlock additional exclusive perks and technical support.

Looking back at October, B.AI reaffirmed the hub value of unified underlying infrastructure in the AI era with a user base nearing 3 million and an ecological matrix of over 70 multimodal foundational models. From the launch of the foundational model rankings to the comprehensive expansion of multimodal computational power, to the minimalist invisible integration experience and the continuous release of inclusive computing resources, B.AI has always been dedicated to folding technological complexity and dismantling computational cost barriers. Looking ahead, B.AI will continue to deepen globalization infrastructure construction, relying on unified API entry points and scaled throughput advantages. Partnering with top global foundational model vendors and a broad community of developers, it will jointly accelerate the comprehensive eruption of next-generation super applications and the Agent economy.