LLM BAZ / NEWS

EDITION 03

EDITION 032026-11-063 STORIES

EDITORIAL NOTE // Model details are attributed to the linked sources. Publisher claims are not independent benchmarks.

THE STORIESSCROLL TO EXPLORE ↓

Local AI

01 / 3

Ollama v0.40.15

More model choices and concurrent requests

Released November 3, Ollama v0.40.15 adds support for MiMo V2.6, which the release notes describe as a model built for code and agent workflows, and AutoTrust decision models. Recurrent and vision GGUF models can now serve more than one request at a time. The notes say autotrust-jev is supported on Apple silicon in this release.

VERSION: v0.40.15
RELEASED: 2026-11-03 (UTC)
NOTE: autotrust-jev supported on Apple silicon in this release

[ RELEASE NOTES ↗ ]

Developer tools

02 / 3

LangGraph 1.2.20

A stricter way to resume interrupted tasks

Released November 5, LangGraph 1.2.20 gives each interrupt in a task its own ID and adds a strict resume mode that refuses stale interrupt answers. For agent workflows that pause for input, this offers a way to reject outdated responses; strict mode is a specific mode, not a claim about all resumes.

VERSION: 1.2.20
RELEASED: 2026-11-05 (UTC)
NOTE: Strict resume mode rejects stale interrupt answers

[ RELEASE NOTES ↗ ]

Infrastructure

03 / 3

LiteLLM v1.106.7

A packaging change for the Lens beta

Released November 4, LiteLLM v1.106.7 moves the Lens beta out of the proxy package and Helm charts. The release notes say deployments with lensWorker.enabled in Helm values must remove that setting before upgrading so the chart renders. The release also adds video endpoint support for registered custom providers. The release notes do not specify the purpose of Lens or further implications of its removal.

VERSION: v1.106.7
RELEASED: 2026-11-04 (UTC)
NOTE: Helm users may need to remove lensWorker.enabled

[ RELEASE NOTES ↗ ]