Baseten Launches Hosted Tools and Grounded Inference for Open-Weight Models

On September 17, Baseten launched the Baseten Hosted Tools ecosystem and the Grounded Inference preview, bringing server-side real-time web search to open/open-weight models for the first time. Built in partnership with Exa and integrating four search partners — Exa, Keenable, p0 and You.com — it can be enabled with a single configuration, with no need for developers to orchestrate tool calls or manage multiple search vendors.

Confirmed

Grounded Inference was co-launched by Baseten and Exa, with Exa's web search as the first available search integration in Hosted Tools.
Four search partners are available under a single configuration: Exa, Keenable, p0 and You.com.
Keenable Search has announced native integration into the Baseten platform; with Grounded Inference, a model can search, read webpages and reason within a single inference request, without stitching together multiple services.
According to @AccBalanced's relaying, the server-side search cuts latency by about 15% and removes orchestration overhead.
Beyond real-time web search, Hosted Tools also covers server-side capabilities such as sandboxed code execution and browser use and automation, targeting open models and agent scenarios.

Why it matters

Previously, developers had to orchestrate search, fetching and reasoning on the client or application layer for open-weight models; Grounded Inference moves these steps server-side and into a single inference request, lowering the barrier to building web-connected applications on open models.
By partnering with four search providers under one configuration, Baseten directly matches the built-in web access of closed-source models.

2026-09-17 ~ 2026-09-18 · 6 related posts

Primary sources

Baseten launches Hosted Tools, bringing server-side web search to open models — baseten · 2026-09-17 05:08 (UTC+08:00)
Baseten partners with Exa to add server-side web search to open-weight models — baseten · 2026-09-18 01:24 (UTC+08:00)
Keenable Search goes native on Baseten: search, read, and reason in one request — baseten · 2026-09-18 02:38 (UTC+08:00)

[source] Baseten launches Hosted Tools, bringing server-side web search to open models — baseten · 2026-09-17
Baseten Adds Server-Side Web Search for Open-Weight Models via One-Line Tool Config — PolarBearby · 2026-09-17
[source] Baseten partners with Exa to add server-side web search to open-weight models — baseten · 2026-09-18
[source] Keenable Search goes native on Baseten: search, read, and reason in one request — baseten · 2026-09-18
Baseten launches server-side web search for open models, claiming 15% lower latency — AccBalanced · 2026-09-18

1 near-duplicate retellings: baseten