Vercel AI Gateway Integrations

Dock serves as a collaborative AI workspace designed for you, your team, and the various agents you deploy. It enables both humans and AI agents to share a unified cloud environment, allowing everyone to access and modify the same information in real-time, rather than navigating through disjointed chats, files, and isolated outputs. The platform is structured around tables with defined columns, rich-text documents, and recognizes agents as primary entities, each equipped with their own API keys, permissions, and audit trails, eliminating the need for delegated human tokens. Teams can leverage Dock for a multitude of tasks, including planning, researching, decision-making, and executing projects, all within a shared interface that accommodates both human and AI contributions. Use cases for Dock span various domains, including engineering, go-to-market strategies, research, operations, individual projects, and agency tasks. Engineering teams can utilize Dock to facilitate sprint planning, create specification documents, and respond to incidents efficiently; marketing teams can streamline content calendars, manage sales pipelines, and enhance customer success initiatives; research teams can effectively document interviews, identify themes, and analyze competitive intelligence; and operations teams can oversee runbooks, manage recruitment processes, ensure compliance, and coordinate onboarding efforts. In essence, Dock fosters a seamless collaboration environment that enhances productivity and innovation across all team functions.

OrcaRouter serves as a routing system for AI models that are compatible with OpenAI, efficiently directing prompts to the appropriate models from a wide array, including OpenAI, Anthropic, Gemini, DeepSeek, Qwen, Kimi, and over 200 other leading and open-source models. Its design aims to maintain the high quality of responses while minimizing costs associated with AI inference by evaluating each prompt and directing complex reasoning tasks to premium models while assigning simpler tasks to more economical open-source options. The routing process is meticulously quality-graded, avoiding arbitrary swaps for cheaper models, and every request clearly indicates the difficulty rating, chosen model, provider, and associated costs, ensuring that routes remain transparent, accountable, and reproducible. Developers can easily switch models by updating the API base URL, while previously established SDKs, model names, and streaming functionalities remain operational. Additionally, OrcaRouter features seamless automatic failover capabilities, allowing for traffic rerouting without interruption should a provider experience downtime, thus preventing disruptions for users. It also offers comprehensive API key management that incorporates spending limits, model allowlists, rate restrictions, and budget compliance, among other functionalities, ensuring robust control over resource usage. This combination of features makes OrcaRouter an indispensable tool for optimizing AI model utilization in various applications.

Sakana Fugu is a multi-agent AI platform and AI model that gives users access to coordinated model intelligence through one API. Instead of relying on one frontier model, Fugu dynamically selects, routes, and coordinates multiple expert models to complete complex tasks more effectively. The system is based on research into learned model orchestration, including the TRINITY and Conductor approaches for assembling agents and guiding collaboration patterns. Fugu is designed for coding, code review, reasoning, research, paper reproduction, cybersecurity analysis, patent investigation, and other work that benefits from multiple specialized agents. Users can access Fugu and Fugu Ultra through an OpenAI-compatible API, making integration easier for existing workflows and developer tools. Fugu is positioned as the default option for everyday use because it balances performance and latency. Fugu Ultra is built for difficult, high-value tasks where maximum quality matters more than speed. The platform also gives organizations the ability to opt out of specific models or providers for data, privacy, compliance, or internal policy reasons. Sakana Fugu helps users reduce dependence on a single AI vendor while gaining a flexible orchestration layer for advanced multi-step AI work.

Wafer is revolutionizing enterprise AI by offering the quickest open-source LLMs, enabling serverless and dedicated inference designed specifically for production workloads. With its serverless inference, teams can utilize top-tier open models without the burden of infrastructure and deployment challenges, providing rapid APIs that include GLM-5.2-Fast for reduced latency through EAGLE speculative decoding and a guaranteed throughput SLA, alongside GLM-5.2, which serves as a flagship model boasting enhanced coding and reasoning abilities. Wafer's innovative technology employs agents to optimize inference throughout the stack, pinpointing and addressing bottlenecks in orchestration, algorithms, serving engines, GPU kernels, and various hardware setups. This system meticulously profiles the stack to determine whether latency or throughput issues arise from factors such as scheduling, decoding, kernels, memory pressure, or hardware compatibility, and then it explores numerous paths to deliver the most effective solution. Rather than depending on a singular switch or heuristic, Wafer undertakes a comprehensive search of combinations involving models, engines, kernels, and hardware to maximize performance. By continually refining these combinations, Wafer ensures that enterprises can operate at peak efficiency while leveraging the best of open-source technologies.

Gemini 3.5 Flash-Lite stands out as the quickest model within Google's Gemini 3.5 lineup, specifically engineered for tasks requiring low latency and for enhancing developer workflows that demand high throughput, including agentic search, document processing, coding, and extensive data analysis. It boasts an impressive output capacity of 350 tokens per second and marks a significant enhancement over earlier Flash-Lite iterations in terms of both quality and agentic capabilities. Developers have the flexibility to adjust the model's thinking level to suit the demands of the task at hand: minimal or low thinking allows for rapid processing of large volumes, while elevated thinking levels accommodate more intricate, multi-step workflows involving subagents. Furthermore, the model is equipped with built-in computational skills, enabling it to interact effectively with various digital environments across compatible platforms. Additionally, Gemini 3.5 Flash-Lite excels in coding, comprehending long contexts, and executing real-world tasks, consistently outperforming its predecessor, Gemini 3.1 Flash-Lite, in critical assessments and even exceeding the performance of Gemini 3 Flash on multiple benchmarks related to agentic functions and software engineering. This impressive performance highlights its potential to transform how developers approach complex workflows and data-intensive tasks.

Prefactor is a cutting-edge platform designed for real-time assessment, monitoring, and reliability of production AI agents. It evaluates each execution instantly based on metrics such as quality, drift, cost, and data risk, seamlessly integrating these assessments into actionable responses to ensure that any failing agent is detected in real time rather than merely reflected on a post-execution dashboard. Teams are equipped to monitor every model invocation, tool usage, and decision-making process through structured traces and spans, allowing them to conduct evaluations using LLM-as-judge, technical assessments, qualitative analyses, and custom metrics at every phase of the process. Additionally, context can be incorporated from various sources, including GitHub, Linear, Jira, databases, and internal APIs, serving as ground truth for evaluations. When a run exceeds predefined limits, Prefactor is capable of blocking or throttling it, pausing sensitive actions, or routing the decision to a person for approval, modification, or rejection prior to execution, with meticulous logging of each choice made. The command-line interface allows for the discovery of agents without the need for platform migration, while the TypeScript and Python SDKs ensure seamless integration with LangChain, Claude, Vercel AI, OpenClaw, and LiveKit, enhancing the overall functionality and adaptability of the platform. This comprehensive approach not only optimizes agent performance but also fosters collaboration among teams by providing clear visibility and control over the AI processes.

MiMo-V2.6-Pro-UltraSpeed is Xiaomi MiMo’s accelerated serving option for MiMo-V2.6-Pro, built for applications that require very high output speed without changing the underlying model quality. Xiaomi states that UltraSpeed can generate output at up to 20 times the speed of the standard MiMo-V2.6-Pro configuration. The model retains MiMo-V2.6-Pro’s natively omnimodal capabilities across coding, agentic workflows, visual reasoning, computer use, and research. Developers can use it for long-horizon software engineering, automation, debugging, tool-driven tasks, and other workloads that benefit from rapid model responses. Its multimodal abilities also support frontend generation, presentation creation, 3D modeling, interactive environments, and visual feedback loops. The broader MiMo-V2.6 architecture combines coding capabilities with 3D spatial reasoning, multimodal perception, and computer-use agent functionality. Xiaomi positions UltraSpeed for real-time interaction and other workflows where response latency is especially important. The accelerated model is offered through MiMo Desktop and can also be called through the Xiaomi MiMo API Platform. MiMo-V2.6-Pro-UltraSpeed is intended for developers and organizations that prioritize maximum generation speed while retaining the capabilities of Xiaomi’s higher-end MiMo-V2.6-Pro model.

Holo4 is H Company's series of generalist computer-use and agentic AI models built to perform multi-step work across software interfaces. It is available as Holo4 27B, a dense 27-billion-parameter model, and Holo4 35B-A3B, a Mixture-of-Experts model containing 35 billion total parameters with 3 billion active. Holo4 can interact with applications by clicking and typing through graphical interfaces, writing and executing code, or calling MCP and API tools. The same model can operate across desktops, websites, Android devices, code sandboxes, and business APIs without requiring developers to select a separate specialized model for each environment. H Company trained Holo4 using 127 billion supervised fine-tuning tokens, with approximately three-quarters consisting of successful agentic trajectories spanning desktop, web, MCP/API, and mobile tasks. Reinforcement learning then trained separate experts for desktop and web interaction and for terminal, MCP, and API work before merging them into a single model. Holo4 27B scored 85.2% on OSWorld, 61.7% on OSWorld 2.0, 45.4% on AutomationBench, and 85.1% on AndroidWorld in the evaluations reported by H Company. The models support a 256K context window, with the 27B model positioned for greater accuracy on long multi-step tasks and the 35B-A3B version positioned as a faster and less expensive alternative. Holo4 is available through a hosted API and downloadable model weights, enabling developers and enterprises to build agents that perform workflows spanning multiple applications and interaction methods.

Recraft is an advanced AI image generation platform built to help designers and creators produce visually appealing content with precision and style. It allows users to generate photorealistic images, vector graphics, and design assets directly from text prompts. One of its standout features is native vector generation, enabling scalable graphics without the need for additional tools. The platform emphasizes strong design quality, delivering outputs that go beyond simple prompt accuracy to include visual taste and consistency. Users can create custom styles by uploading reference images, which can then be reused across projects. Recraft also includes a suite of editing tools such as background removal, image upscaling, and object editing. It supports a variety of use cases, including logos, ads, mockups, and social media visuals. The platform is designed to streamline creative workflows and reduce the need for multiple design tools. Its intuitive interface makes it accessible to both professionals and beginners. By combining generation and editing in one place, it simplifies the content creation process. Ultimately, Recraft enables users to produce high-quality, consistent visuals at scale.

Qwen3.6-Plus is a state-of-the-art AI model designed to support real-world agentic applications, advanced coding, and multimodal reasoning. Developed by the Qwen team under Alibaba Cloud, it offers a significant upgrade over previous versions with improved performance across coding, reasoning, and tool usage tasks. The model features a 1 million token context window, enabling it to handle long and complex workflows with high accuracy. It excels in agentic coding scenarios, including debugging, repository-level problem solving, and automated development tasks. Qwen3.6-Plus integrates reasoning, memory, and execution into a unified system, allowing it to operate as a highly capable autonomous agent. Its multimodal capabilities enable it to process and analyze text, images, videos, and documents for deeper insights. The model supports real-time tool usage and long-horizon planning, making it ideal for enterprise and developer use cases. It is accessible via API through Alibaba Cloud Model Studio and integrates with popular coding tools and assistants. Developers can leverage features like preserved reasoning context to improve performance in multi-step tasks. Overall, Qwen3.6-Plus empowers businesses and developers to build intelligent, scalable, and autonomous AI-driven applications.

Xiaomi MiMo-V2.5-Pro is a next-generation open-source AI model designed for advanced reasoning, coding, and long-horizon task execution. It uses a Mixture-of-Experts architecture with over one trillion parameters and a large active parameter set for efficient performance. The model supports an extended context window of up to one million tokens, allowing it to handle complex, multi-step workflows. It is built to perform autonomous tasks, including software development, system design, and engineering optimization. Benchmark results show strong performance across coding, reasoning, and agent-based evaluation tests. MiMo-V2.5-Pro incorporates hybrid attention mechanisms to improve efficiency while maintaining accuracy across long contexts. It is optimized for token efficiency, reducing the computational cost of running complex tasks. The model can integrate with development tools and frameworks to support real-world applications. It is designed to complete tasks that would typically require significant human effort over extended periods. Xiaomi has made the model open source, enabling developers to access and customize it. By combining performance, scalability, and efficiency, MiMo-V2.5-Pro pushes the boundaries of modern AI capabilities.

Xiaomi MiMo-V2.5 is a next-generation open-source AI model that combines agentic intelligence with multimodal capabilities. It is designed to process and understand text, images, and audio within a single architecture. The model uses a sparse Mixture-of-Experts framework with a large parameter count to deliver efficient and scalable performance. It supports a context window of up to one million tokens, allowing it to handle long and complex workflows. MiMo-V2.5 integrates visual and audio encoders to improve perception and cross-modal reasoning. It is capable of performing tasks such as coding, reasoning, and multimodal analysis with strong accuracy. Benchmark results show competitive performance compared to leading AI models in both agentic and multimodal tasks. The model is optimized for token efficiency, balancing performance with lower computational cost. It is designed for real-world applications that require both reasoning and perception. Xiaomi has open-sourced the model, making it accessible for developers and researchers. By combining multimodality, scalability, and efficiency, MiMo-V2.5 pushes forward the development of advanced AI systems.

Qwen3.7-Plus is an advanced multimodal agent model that seamlessly integrates vision and language into a single, adaptable foundation for intelligent agents. Expanding upon the agentic intelligence of Qwen3.7, it enhances its abilities to include visual comprehension, reasoning, grounded interactions, and the use of various multimodal tools, allowing agents to perceive, analyze, and operate within text, images, documents, screens, and intricate real-world scenarios. This model is specifically crafted for dynamic tasks that go beyond mere static question answering, facilitating activities such as visual searches, document understanding, chart and table evaluations, screen comprehension, GUI interactions, image-driven reasoning, and workflows where perception, planning, and action are interlinked. Qwen3.7-Plus fortifies the relationship between linguistic reasoning and visual cues, empowering users to inquire about images, decode complex multimodal information, extract organized data, and formulate responses that incorporate both contextual and visual elements, thus broadening the scope of interactive AI applications. With these enhancements, users can engage in more sophisticated and nuanced interactions with the system, making it a powerful tool for various practical applications.

Grok Voice Think Fast 2.0 stands as the premier voice model from xAI, designed for the creation of real-time assistants, telephone agents, and interactive voice systems capable of bidirectional audio and text streaming via WebSocket. Developers have the flexibility to tailor various system parameters, such as the level of reasoning effort, the choice between built-in or custom voices, automatic voice activity detection on the server side, as well as configurable settings for silence duration, idle re-engagement, playback speed, and the ability to resume sessions following temporary disconnections. The model processes audio in several formats, including PCM, G.711 μ-law, G.711 A-law, and Opus, accepting both JSON and raw binary frames, with the adaptability to adjust PCM sample rates ranging from standard telephone quality to 48 kHz. It boasts support for over 20 languages with native-like accents, features automatic language recognition, generates natural responses in the user's preferred language, and facilitates smooth code-switching. Additionally, the inclusion of language hints and the ability to incorporate up to 100 key terms significantly enhance the accuracy of transcribing regional dialects, names, product identifiers, codes, addresses, and other specialized vocabulary, while pronunciation adjustments ensure the spoken output is correct and intelligible. This versatility makes Grok Voice Think Fast 2.0 an invaluable tool for developers looking to enhance user interaction through voice technology.

The new OpenAI API service tier, GPT-5.6 Sol Ultrafast, operates up to 14 times quicker than the Standard processing version, delivering cutting-edge intelligence to applications and workflows where every fleeting moment is crucial. Utilizing Cerebras technology, it boasts the capability to produce as many as 750 output tokens each second, enabling sophisticated reasoning to function at real-time velocities without the need for a more compact or specialized model. This service is particularly tailored for business environments where rapid responses can significantly enhance the capabilities of AI systems. It has various applications, including incident response, where it can swiftly analyze logs, code changes, traces, and engineering reports during ongoing outages; financial research and security, where it can rapidly evaluate fluctuating market signals and identify suspicious transactions; and customer support, where intricate problems can be resolved seamlessly during live conversations. In the realm of e-commerce, it excels at handling product inquiries, verifying inventory status, and customizing product recommendations to enhance user experience. By implementing this advanced service, organizations can expect improved efficiency and effectiveness in their operations.

Gemini 3.8 Flash Cyber represents Google's most advanced cybersecurity model, offering top-tier performance in identifying vulnerabilities and automating patching processes with remarkable speed for rapid iteration. Tailored for trusted defenders, it is accessible via the Fairwind Program. On CyberGym, a recognized industry benchmark for detecting vulnerabilities, this model showcases exceptional autonomous vulnerability discovery, outperforming both Gemini 3.5 Flash Cyber and larger frontier models. Furthermore, Google assessed its effectiveness on an internal benchmark that spans complex codebases across 20 programming languages, achieving a success rate of over 70% in identifying various vulnerabilities. Unlike many models that focus on offensive strategies, Gemini 3.8 Flash Cyber emphasizes the importance of fixing vulnerabilities, providing defenders with advanced tools that enhance their ability to stay ahead of cyber attackers. This focus on proactive defense represents a crucial shift in the cybersecurity landscape, prioritizing the safeguarding of systems over mere exploitation capabilities.

OpenAI's Images 2.5 is their cutting-edge image model, delivering enhanced detail, improved editing accuracy, quicker generation times, and superior tools for visual creation and refinement. It achieves more lifelike lighting and richer textures, ensures subjects in reference images are maintained more consistently, and adheres to editing requests with greater precision over multiple interactions. The reduction in generation latency by as much as 50% compared to Images 2.0 enables users to rapidly iterate their concepts. This model excels at implementing focused modifications to individual elements while preserving the integrity of the subject, composition, background, and surrounding details. Furthermore, during extended editing discussions, earlier revisions are more likely to stay intact, preventing any degradation in image quality over time. Images 2.5 also significantly enhances the understanding of intricate visual instructions, real-world contexts, artistic styles, transparent backgrounds, layouts, and complex compositions, facilitating a more intuitive creative process. Ultimately, this advancement allows for a more seamless and dynamic user experience in visual editing.

Seedance 2.0 is a next-generation AI video creation model developed by ByteDance to simplify high-quality video production. It allows users to generate complete videos using text, images, audio, and existing clips as creative inputs. The platform excels at maintaining visual coherence, ensuring characters, styles, and scenes remain consistent across shots. Advanced motion synthesis enables smooth transitions and realistic camera movement throughout each video. Users can reference multiple assets at once, combining visuals and sound to shape the final output. Seedance 2.0 removes the need for traditional editing tools by handling pacing and shot composition automatically. Videos are produced in professional-grade resolutions suitable for commercial use. The model has gained attention for producing complex animated sequences, including anime-style visuals. It empowers individual creators and small teams to achieve studio-like results. At the same time, it introduces new conversations around responsible AI use and content authenticity.

ChatGPT Images 2.0 is an advanced AI-powered image generation model created by OpenAI to deliver more accurate and practical visual outputs. It introduces a reasoning-based approach, allowing the system to plan and interpret prompts before generating images. This results in improved accuracy, better composition, and more consistent visual details. The platform excels at rendering text within images, supporting multilingual typography with high precision. It can generate multiple related images from a single prompt while maintaining consistency across characters and scenes. The model supports higher resolutions and flexible aspect ratios, making it suitable for professional use cases. ChatGPT Images 2.0 is designed for real-world applications such as marketing, presentations, storyboards, and product visuals. It also integrates with ChatGPT, making image creation part of a broader workflow. Compared to earlier versions, it provides more reliable outputs with fewer distortions or errors. The system can handle complex layouts, including infographics and UI designs. By combining reasoning, accuracy, and flexibility, ChatGPT Images 2.0 represents a major step forward in AI-generated visuals.

Grok Voice Think Fast 1.0 is a next-generation voice AI model from xAI that is built to manage complex, multi-step conversational workflows in real-world environments. It is designed for use cases such as customer support, sales, and enterprise automation, where accuracy and speed are critical. The model delivers fast, natural-sounding responses while performing real-time reasoning in the background without increasing latency. It can handle ambiguous requests, interruptions, and diverse accents, making it highly effective in real-world voice interactions. Grok Voice excels at structured data collection, accurately capturing details like phone numbers, addresses, and account information. It supports over 25 languages, enabling global deployment across different markets. The model is optimized for high-volume tool usage, allowing it to interact with multiple systems during a conversation. It has been tested in challenging environments, including noisy telephony scenarios. Its strong reasoning capabilities help reduce errors and improve response reliability. Overall, it empowers organizations to automate complex voice-based workflows with confidence and efficiency.

Claude Fable 5.5 is an anticipated but currently unannounced model in Anthropic's Claude family, and Anthropic has not confirmed that a model with this name will be released. As of September 30, 2026, Claude Fable 5.1 remains the latest officially documented Fable model. Fable represents Anthropic's highest-end model tier for demanding reasoning and long-horizon agentic work, while the newer Opus 5.5 and Sonnet 5.5 occupy lower-cost positions in the Claude lineup. Anthropic's current documentation gives Fable 5.1 a 1-million-token context window and maximum output length of 128,000 tokens. It supports text and image inputs with text output and uses adaptive thinking that remains active throughout model operation. Fable 5.1 defaults to high reasoning effort and is listed as having a June 2026 reliable knowledge cutoff and training-data cutoff. API pricing is $10 per million input tokens and $50 per million output tokens, while prompt-cache reads cost $0.25 per million tokens and Batch API processing receives a 50% input and output discount. Anthropic's official documentation currently provides model identifiers for Fable 5.1 across the Claude API, Amazon Bedrock, Google Cloud, Microsoft Foundry, and Claude Platform on AWS. No equivalent model identifier, specifications, benchmark results, pricing, availability information, or release schedule has been published for Claude Fable 5.5.