Inkling 2 Small Lands on AI Gateway: A New Multimodal Coding Model Option

Thinking Machines' multimodal reasoning model Inkling 2 Small lands on Vercel AI Gateway, supporting coding agents and visual iteration.
Thinking Machines' Inkling 2 Small is now on Vercel AI Gateway—a multimodal reasoning model with text and image input and text output, aimed at coding agents, visual understanding, and professional knowledge work. It handles long coding tasks, calls tools to gather information and produce documents and spreadsheets, reads text from images, interprets charts, and uses screenshots to verify code output in an iterative loop. Developers call it via the model name thinkingmachines/inkling-2-small with AI SDK's streamText, or set it up in Claude Code, Codex, and Cursor via one-click Vercel CLI configuration. Gateway passes through vendor pricing with no markup and offers retries, failover, and usage tracking—ideal for teams needing unified multi-model orchestration.
Vercel AI Gateway has recently added the Inkling 2 Small model from Thinking Machines. It's a multimodal reasoning model that supports text and image input with text output, positioned for scenarios including coding agents, visual understanding, and professional knowledge work. For developers building agent workflows or looking to orchestrate models through a unified interface, this launch offers a new option.

Model Capabilities: From Writing Code to Delivering Documents
Inkling 2 Small's coding ability goes beyond generating code snippets. According to the official description, it can handle longer coding tasks, call tools to gather information, and turn results into deliverables such as documents and spreadsheets. This means it's better suited for being placed inside an agent loop rather than for one-off Q&A.
Visual capability is another highlight. The model can read text from images, interpret charts and infographics, and even inspect interfaces. In visual development scenarios, it can use screenshots to verify the code output it produced and guide the next round of revisions accordingly—this "generate-verify-iterate" loop is exactly one of the capabilities today's coding agents need most.
Agent loop refers to the operating mode in which a model autonomously completes multi-step tasks in a loop of "observe—decide—act—observe again": the model first analyzes the current state, chooses to call a tool (retrieval, code execution, file read/write, etc.), reads the tool's returned result, and then decides the next step. Unlike one-off Q&A, a model in an agent loop needs to maintain task state, plan subgoals, and self-correct on erroneous results. Therefore, evaluating whether a model suits agent scenarios often weighs instruction retention over long contexts, stability of tool-call formatting, and the ability to adjust strategy after multiple rounds of failure—not just the code quality of a single-turn generation.
Integration: Unified Model Name and One-Click CLI Setup
Use the model name thinkingmachines/inkling-2-small when calling. The official AI SDK example uses streamText to send a streaming request directly—pass in the model name and prompt to get a text stream, which is great for quick validation.
To use it in coding tools like Claude Code, Codex, and Cursor, install the latest Vercel CLI and run npx vercel ai-gateway setup, then select the model in your agent. For developers already accustomed to switching models in these tools, the configuration cost is low.
AI SDK is an open-source TypeScript toolkit from Vercel for calling large models in JavaScript/TypeScript applications. It abstracts different vendors' APIs into a unified interface, with core functions including generateText (one-shot generation) and streamText (streaming generation, returning iterable text increments). Developers only need to swap the model identifier to switch vendors within the same codebase, without rewriting request logic. AI SDK also supports tool calling, structured output, and multi-step agents—which is why it's often used as the foundation for agent workflows.
The Value of AI Gateway: More Than a Model List
AI Gateway is positioned as a unified API layer: calling models, tracking usage and cost, configuring retries, failover, and performance optimization to achieve higher availability than any single vendor. It also includes custom reports, zero data retention support, API key budgets, and routing rules.
Pricing is worth noting: Gateway passes through vendor pricing as-is, with no markup and no platform fees for inference—the same applies to BYOK (bring your own key) requests. For teams that need multi-model comparison or are cost-sensitive, this no-markup intermediary layer lowers the barrier to experimentation.
BYOK (Bring Your Own Key) means developers use API keys they've obtained from model vendors and route requests through the gateway. Billing still happens directly between the developer and the vendor; the gateway only handles routing, observability, and policy enforcement, without involving financial settlement. The value of BYOK is that teams can use their existing contract prices and quotas while enjoying a unified interface, usage tracking, and failover; for the gateway provider, it reduces compliance and credit risks associated with payment processing and resale. The opposite is the gateway-paid model, where the platform purchases centrally and bills by usage.
Who Should Try It
Professional knowledge workers who need multimodal input, teams building coding agents, and developers who want unified model orchestration without vendor lock-in can all add Inkling 2 Small to their candidate list. It's recommended to first run typical tasks in the playground, then evaluate its actual performance on long tasks and tool calling.
Related articles

AI Writes Your Code—Do You Still Need to Learn Git? Five Reasons
AI coding agents handle Git operations automatically—is it still necessary to learn Git? Five reasons: team communication, reviewing agent PRs, monitoring code changes, fallback when AI goes down, and the joy of learning—plus how the focus of learning Git is shifting in the agent era.

Anthropic Reportedly Prices IPO at $83 Per Share, Valuing Company at $1.5 Trillion
Anthropic reportedly prices IPO at $83 per share, above the expected range, with a fully diluted valuation of $1.5 trillion. This article analyzes the signal significance of the pricing and its impact on the AI industry's financing environment.

Open-Source Code Hosting Platform Codeberg Suffers Outage, Sparking Debate in Developer Community
Open-source code hosting platform Codeberg.org suffered an outage, sparking Hacker News debate over nonprofit platform stability and open-source infrastructure dependency risks.