AWS、Cloudflare Databricks reportedly follows with decision models

Article

AWS、Cloudflare Databricks is also following Jev, and decision models are quickly becoming standard

Beating AI News Flash: After Jev was released,AWS、Cloudflare and Databricks have recently also followed with decision models.AWS launched the open-source Strands Decider 2B. It is based on Qwen3.5-2B, removes the text-generation module, replaces it with a decision module dedicated to selecting answers, and then fine-tunes it with LoRA。In official tests, its median latency for a single decision on RTX 3090 was approximately 115ms, with complete training taking approximately 11 hours.Cloudflare released Clef and Clef-flash at once, based respectively on Qwen3.8-27B and Qwen3.5-9B and making both fully open source. In Cloudflare tests on Jev's official evaluation set, the Clef series exceeded Jev on 4-type tasks in terms of 3-type, while Clef-flash had a median latency of 38.8ms, compared with 524.1ms for Jev.

Databricks has integrated this capability directly into its platform. Its newly launched ai_decide can be called from SQL or REST API to classify and score data and perform Agent routing; the interface is also directly compatible with the TypeSafe AI API. The company says a single decision can be returned in under 1 seconds.

From developers replicating the approach themselves to AWS、Cloudflare and Databricks following suit, decision models are rapidly changing from a new concept into a common component of Agent infrastructure.

Original link https://m.theblockbeats.info/flash/370133

Key points

The article stated that after Jev was released,AWS、Cloudflare and Databricks followed with decision models.
AWS launched the open-source Strands Decider 2B, based on Qwen3.5-2B, and using a dedicated decision module and LoRA for fine-tuning.
The article stated that Strands Decider 2B has a median latency of approximately RTX 3090 on 115ms per single decision, with complete training requiring approximately 11 hours.
Cloudflare released Clef and Clef-flash, based respectively on Qwen3.8-27B and Qwen3.5-9B, and making both fully open source.
The article stated that in Cloudflare tests on Jev's official evaluation set, the Clef series exceeded Jev on 4-type tasks in terms of 3-type, while Clef-flash had a median latency of 38.8ms, compared with 524.1ms。
Databricks integrated ai_decide into its platform, enabling classification, scoring, and SQL through REST API or Agent routing, and providing compatibility with TypeSafe AI API。
The article stated that Databricks' single decisions could be returned in under 1 seconds.

AI-assisted interpretation

The following is analysis, separate from reported facts. Verify important claims independently.

Decision models primarily select from multiple answers or actions rather than generating long-form text. The products mentioned in the article can be used for classification, scoring, and Agent routing, but the related release and testing information still requires further verification.

Why it matters to readers

If these products and performance figures are verified, decision models may become Agent more common foundational components in systems.

This indicates that AI applications may need not only to generate content, but also a dedicated module to determine what to do next.

Risks and unknowns

The evidence comes from a news channel and does not provide AWS、Cloudflare or first-party announcements from Databricks.
The product releases, evaluation results, and latency figures in the article cannot currently be independently verified.
The evaluation set, testing methodology, and latency measurement criteria are not specified in the evidence.

AWS、Cloudflare Databricks reportedly follows with decision models

Article

AI-assisted interpretation

Why it matters to readers

Risks and unknowns

Related Developments

Related concepts

Keep following

Security Center