Vercel AI SDK Adds Support for Gemini Embedding 2 and Deep Research Models
Vercel released version 4.0.0-canary.75 of its AI SDK Google package on May 30, adding support for three new Google models: gemini-embedding-2, deep-research-max-preview-04-2026, and deep-research-preview-04-2026. The update enables developers to integrate Google's latest embedding and deep research capabilities into applications built with the Vercel AI SDK.
Vercel AI SDK Adds Support for Gemini Embedding 2 and Deep Research Models
Vercel released version 4.0.0-canary.75 of its AI SDK Google package on May 30, 2025, adding support for three new Google models: gemini-embedding-2, deep-research-max-preview-04-2026, and deep-research-preview-04-2026.
The update is a patch release in the 4.0.0 canary series and includes support for both the standard Google AI and Google Vertex AI implementations.
New Models
The three models added in this release:
- gemini-embedding-2: Google's second-generation embedding model for text representation and semantic search
- deep-research-max-preview-04-2026: A preview version of Google's deep research model, with the "max" designation suggesting enhanced capabilities
- deep-research-preview-04-2026: The standard preview version of Google's deep research model
The "04-2026" version tag indicates these are April 2026 preview snapshots, though the naming convention suggests they are forward-dated model identifiers rather than current releases.
Google has not publicly announced these deep research models, and the "preview" designation in the model names indicates they are early-access versions. Pricing and benchmark data are not yet disclosed.
SDK Integration
The Vercel AI SDK provides a unified interface for developers to integrate multiple AI providers into their applications. This update allows developers using the SDK to access Google's latest models through both direct Google AI API calls and Google Cloud's Vertex AI platform.
The canary release tag indicates this is a pre-release version intended for testing and early adoption, not yet recommended for production use.
What This Means
This SDK update reveals Google is developing dedicated "deep research" models, likely designed for complex analytical tasks requiring extended reasoning or comprehensive information synthesis. The dual-tier approach (standard and "max") mirrors Google's existing Gemini lineup strategy. Developers building with the Vercel AI SDK gain early access to these capabilities, though production deployment should wait for stable releases with published pricing and performance metrics.
Related Articles
Vercel AI SDK Adds Support for Google's Deep Research Models and Gemini Embedding-2
Vercel released version 5.0.0-canary.98 of its Google Vertex AI SDK, adding support for three new Google models: deep-research-max-preview-04-2026, deep-research-preview-04-2026, and gemini-embedding-2. The update enables developers to integrate Google's research-focused models and latest embedding model into applications using Vercel's AI SDK.
Google preparing 'AI Ultra Lite' tier between $20 Pro and $250 Ultra plans, adding usage dashboard
Google is developing an intermediate subscription tier called 'AI Ultra Lite' to slot between its $20 Pro and $250 Ultra plans, according to code discovered in the Gemini macOS app. The company is also preparing a usage dashboard showing token budgets across five-hour and weekly limits.
Vercel AI SDK Deprecates searchParameters for xAI, Adds Image Search Support
Vercel released AI SDK version 4.0.0-canary.69 with breaking changes to xAI integration. The update deprecates the searchParameters option for xAI live search, replacing it with dedicated web_search and x_search agent tools, and adds image search capability through a new enableImageSearch parameter.
Anthropic Python SDK v0.104.0 adds thinking token count estimates for streaming responses
Anthropic released version 0.104.0 of its Python SDK on May 21, 2026. The update adds support for a thinking-token-count beta feature that provides estimated token counts in thinking block deltas when streaming responses from reasoning models.
Comments
Loading...