AI Tool Updates - September 25, 2026
This period's AI tooling news is dominated by incremental shipping: Ollama's release candidate makes MLX the default runner on Apple Silicon, GitHub moved to default-enable generally available…
AI Tool Updates - September 25, 2026
Week of: September 25, 2026
Overview
This period's AI tooling news is dominated by incremental shipping: Ollama's release candidate makes MLX the default runner on Apple Silicon, GitHub moved to default-enable generally available Copilot features for Business and Enterprise customers, and CodeQL 2.27.1 expanded its query and language coverage. SDK and workflow platforms also pushed patch releases, with the Vercel AI SDK extending provider options for xAI and n8n and LangChain shipping maintenance updates.
Stories
1. Ollama v0.40.0-rc0 runs supported models on MLX by default on Apple Silicon
Source: Ollama Link: https://github.com/ollama/ollama/releases/tag/v0.40.0-rc0
Ollama published v0.40.0-rc0, in which model architectures supported by the MLX runner run by default on Apple Silicon devices. The release notes illustrate usage with ollama pull qwen3.8 and ollama run qwen3.8, and state that additional models will be tested and enabled during the RC period. The changelog spans v0.34.4 to v0.40.0-rc0.
Local inference performance on Apple hardware has been a practical constraint for developers running models on laptops, and making MLX the default path means Mac users get the optimized backend without opting in. Teams shipping local-first agents or copilots should validate their supported model list against the MLX runner before the RC graduates.
Impact Analysis: Mac-based AI development gets a faster default inference path, but the RC status means teams should pin versions and test model compatibility rather than relying on it in production.
2. GitHub introduces a global default policy for Copilot features in Business and Enterprise
Source: GitHub Changelog Link: https://github.blog/changelog/2026-09-24-default-enablement-of-copilot-features-for-copilot-business-and-enterprise
GitHub announced a new global default policy governing generally available GitHub Copilot features and supported client capabilities within enterprise and organization Copilot settings. The changelog notes a 28-day window during which administrators can act, with details covered in the post (see source).
Default-enablement shifts the administrative burden from opting in to opting out, which changes how quickly new Copilot capabilities reach developer desktops in large organizations. Administrators with compliance or data-handling constraints will need to review enterprise and organization settings during the stated window rather than after rollout.
Impact Analysis: Enterprise Copilot admins should audit their organization and enterprise policies now, since new capabilities will arrive enabled by default unless explicitly disabled.
3. CodeQL 2.27.1 adds C and C++ queries, C# queries, and Kotlin 2.4.20 support
Source: GitHub Changelog Link: https://github.blog/changelog/2026-09-25-codeql-2-27-1-adds-c-and-c-query-and-kotlin-2-4-20-support
CodeQL 2.27.1 introduces new queries for C/C++ and C#, support for Kotlin 2.4.20, and what the changelog describes as query-accuracy improvements. CodeQL is the static analysis engine behind GitHub code scanning.
Language and query coverage directly affect how much code scanning can catch in mixed-language repositories, and Kotlin version support matters for Android and JVM teams whose toolchains move quickly. For teams running AI-generated code through CI, tighter static analysis is one of the few automated gates that can catch issues before merge.
Impact Analysis: Upgrading CodeQL buys broader language coverage and fewer false positives, making code scanning a more reliable gate for AI-assisted contributions.
4. Vercel AI SDK ships xAI provider options and coordinated dependency bumps
Source: Vercel AI SDK Link: https://github.com/vercel/ai/releases/tag/%40ai-sdk%2Fxai%405.0.8
@ai-sdk/xai@5.0.8 adds missing video provider options and missing Responses API provider options. In the same release wave, packages including @ai-sdk/workflow@2.0.45, @ai-sdk/vue@4.0.114, @ai-sdk/svelte@5.0.114, @ai-sdk/tui@1.0.115, and @ai-sdk/workflow-harness@1.0.124 were published as patch releases updating their dependencies.
The additions close gaps in the xAI provider surface for video and the Responses API, which matters for builders wrapping multiple model providers behind a single interface. The coordinated patch set across framework adapters is routine maintenance, but version alignment across the SDK's many packages is a recurring integration concern.
Impact Analysis: Teams using the xAI provider should upgrade to pick up video and Responses API options; others can treat the remaining packages as routine dependency bumps.
5. n8n publishes patch releases across its 2.x and 1.x lines
Source: n8n Link: https://github.com/n8n-io/n8n/releases/tag/n8n%402.41.3
n8n released 2.41.3 and 2.40.7 on September 25, both containing a core bug fix that propagates project span attributes to node spans, plus 1.123.82 on the older 1.x line, which bumps adm-zip to 0.6.1. The beta and stable release tags point at 2.41.3 and 2.40.7 respectively.
For teams running agentic or automation workflows on n8n, correct span-attribute propagation improves the observability of individual nodes within a workflow execution. The adm-zip bump on the maintenance branch is a dependency update on a line that is no longer receiving feature work.
Impact Analysis: Self-hosted n8n operators should upgrade to restore accurate per-node tracing, and 1.x users should plan a migration rather than wait for further fixes.
6. langchain-fireworks 1.6.3 released
Source: LangChain Link: https://github.com/langchain-ai/langchain/releases/tag/langchain-fireworks%3D%3D1.6.3
LangChain published langchain-fireworks 1.6.3, with changes since 1.6.2 referencing several pull requests (#40834, #40814, #40818, #40804). The release notes list only PR numbers without descriptive text.
A patch bump on a single provider integration package is minor in isolation, but it is the kind of release that can ripple through pinned dependency sets in agent applications. Teams tracking LangChain's provider packages should consult the linked pull requests for specifics, since the changelog itself carries no detail.
Impact Analysis: Low-impact maintenance release; check the referenced PRs if you depend on the Fireworks integration, otherwise upgrade on your normal cadence.
Source Links
More from News
