Sage 2 Brings Evidence-Backed Extraction

// 1h agoMODEL RELEASE

Sage 2 Brings Evidence-Backed Extraction

Levanto Labs’ Sage 2 lets developers define fields and receive structured values with source text, making extraction more traceable and reliable. It extends the System One approach beyond decisions into machine-readable data capture.

// ANALYSIS

Sage 2 could become the missing companion to decision models like Jev: extract facts first, then decide what to do with them.

–Typed outputs reduce fragile JSON parsing and post-processing
–Source text gives developers an evidence trail for audits and human review
–Extraction plus calibrated decisions creates a practical pipeline for agents and workflows
–The key test is whether abstention and provenance remain reliable on messy, ambiguous documents
–Closed, hosted models will need strong latency, pricing, and benchmark transparency to win adoption

// TAGS

sage-2structured-outputdata-toolsapiagentllm

DISCOVERED

1h ago

2026-11-02

PUBLISHED

2h ago

2026-11-02

RELEVANCE

9/ 10

AUTHOR

omarsar0

// KEEP READING

More AI developer news from the feed

EXPLORE FULL FEED

UPDATE32m ago

Anthropic Messages API Adds Server-Side Threads

Anthropic’s Messages API is adding public-beta message threads that keep conversation state on the server. Developers can continue threads with a previous message ID and send only new turns, reducing repeated context transfer for long sessions.

UPDATE47m ago

Revyl Agents Test iPhone Duo Layouts

Revyl now lets coding agents test and fix iOS app layouts for Apple’s foldable iPhone Duo. The update extends its agentic mobile-testing workflow to Duo-specific screen configurations before release.

BENCHMARK53m ago

ElevenAgents Finds Voice Wins Complex Calls

ElevenLabs analyzed 469 production agents and found that voice outperformed chat as conversations became longer and more complex. For calls over two minutes, voice achieved 31.7% higher success rates and resolved successful interactions roughly one-third faster.