FR
live
tag

#open-source

Jais 3 ships six Arabic-native open models, from phone-size to 375B

On October 15, 2026, the Institute of Foundation Models (MBZUAI) and Inception42 released Jais 3, a fleet of six open models built for Arabic — from 0.9B to 375B parameters — paired with a speech recognition model for the region’s dialects. For organizations that work in Arabic, it is now the sovereign baseline to test before any Western general-purpose model.

The Dutch tax office abandons Microsoft 365 and brings email and calendars on-premises in 2027

On October 7, 2026, the Dutch State Secretary for Finance announced that the Tax and Customs Administration is dropping its Microsoft 365 migration: email and calendars move on-premises in 2027, with European open source for storage and collaboration to follow. For European IT leaders, it is the most concrete signal yet that sovereignty is no longer a slogan but a budgeted trajectory.

Cloudflare and AWS open the open-weight decision-model lane for agents

On 1 October 2026 Cloudflare released Clef and Clef-flash, and AWS Strands Decider 2B: open-source decision models that answer with typed probabilities instead of text, built for agent decision-making. They undercut the closed Jev model and raise the question of what you should stop sending to the frontier model.

Olmo-core 3 opens MoE training to the trillion-parameter scale

Ai2 ships Olmo-core 3, an open mixture-of-experts training stack that holds a 1.2-trillion-parameter model across 512 GPUs. The FSDP-to-DDP switch and MXFP8 precision change the compute economics for labs that do not have Megatron-Core.

The self-hosted community unmasks the developer of BookOrbit

On September 18, 2026, the self-hosted community identified the developer of BookOrbit — a self-hosted reading platform with about 4,600 GitHub stars — as the creator of Booklore, a twin project that disappeared after being called out for code quality, contributor mistreatment, and license abuse. Before deploying software you run yourself, verify author continuity, the license, and community history.

IBM and NASA release an open-source foundation model for mapping the Moon

IBM and NASA open-source the NASA-IBM Lunar Foundation Model, a foundation model trained on decades of multi-instrument lunar data, paired with a unified 30-layer dataset. It cuts ice-deposit identification error by up to 22% and joins the Prithvi family.

Papermerge shifts its focus to SaaS, its open-source edition hunting for maintainers

After nine years of open-source development, the Papermerge document management system is splitting in two: a SaaS launching on November 11, 2026, and a self-hosted edition whose author is now “seeking maintainers” while he transitions to the cloud. If your archive runs on Papermerge, plan a migration instead of waiting to see what happens next.

A Void Linux maintainer orphans 113 packages after an AI-policy dispute

On September 12, 2026, contributor Andrea Brancaleoni orphaned 113 Void Linux packages — including Kubernetes, Docker, Terraform, and Thermald — after being called out for using an LLM in an undisclosed update comment. AI policies are becoming a maintenance-continuity problem as much as a contribution-ethics one.

SourceHut Bans LLM-Generated Code, Following Codeberg

SourceHut changes its terms of service to prohibit LLM-assisted code, becoming the second major forge to follow Codeberg’s lead. If your project relies on generative AI tooling, prepare a migration to Forgejo or a self-hosted instance.

AWS acquires DuckLabs, the company behind DuckDB, and keeps the project MIT-licensed under independent governance

Announced on August 26, 2026 and confirmed in the August 31 AWS weekly roundup, the acquisition of DuckLabs by AWS brings the DuckDB maker in as a subsidiary, without touching the MIT license or the DuckDB Foundation governance. For data teams, DuckDB remains a safe bet; the open question is whether the server-side roadmap tilts toward AWS.

Hugging Face separates open-model attention from actual adoption

Hugging Face’s summer 2026 report shows that media attention and real adoption of open models barely overlap anymore, and that Chinese labs dominate the frontier by sheer size. Small models and Qwen remain the practical layer, while agents become the Hub’s primary user.

Debian puts LLM use in its contributions to a project-wide, eight-option vote

From August 15 through August 28, 2026, Debian Developers are voting on a General Resolution governing LLM use in the project’s contributions, with eight proposals and a “None of the above” option. The outcome will set a de facto standard for the supply chain of enterprise Linux distributions.

Qwen3.8-27B ships a 27-billion-parameter multimodal model under Apache 2.0

On August 14, 2026 Alibaba’s Qwen team released Qwen3.8-27B, a dense 27-billion-parameter multimodal model under an Apache 2.0 license that beats larger models on agentic coding. For teams self-hosting their models, it is a serious candidate to replace proprietary APIs on development tasks.

Type at least two characters.

↑ ↓ navigate ↵ open esc dismiss