#Open-source ecosystem

AIReJev, an independent community project, applied LoRA post-training to OpenBMB's MiniCPM5-2B for bounded agent decisions: state, question, and candidate options yield one choice. On its sealed 1,892-sample holdout, accuracy rose from 51.11% to 80.50% (+29.39 percentage points) with 0% invalid outputs, at about $5.31 in cumulative Modal billing including earlier experimental overhead. The authors describe this as an early, task-specific result, not parity with Jev.

AIDefenders can’t learn from what they can’t see. If you have traces and are being pressured to keep them private, my DMs are open. Let’s level the playing field and fight the asymmetry and lack of transparency in AI!

AIStepFun's Step 5 Preview is now live on OpenRouter, with a week of free access rolling out across opencode, Cline, NousResearch, and Kilo Code. The company positions it as flagship-tier intelligence for agentic and professional work at substantially lower task cost, letting users keep their workflow while switching models.

AIYes, open models are being fitted to be used by some attackers, but so long as Claude Code can be the orchestrator of a hack we desperately need open models to diffuse defense as well. A hard balance societally to stomach, but needed.

AIIBM's torch-spyre integration makes Spyre, its dataflow inference accelerator, a native PyTorch device by mapping PyTorch's device, allocator, stream, and event abstractions onto the Spyre runtime and firmware. Tensors stay resident on device="spyre" between operations, and FX graphs remain in the Inductor compiler path. The approach gives eager and compiled execution one path with lower launch overhead.

AIJetBrains released Mellum2.1, a 12B mixture-of-experts model with 2.5B active parameters under the Apache 2.0 license, built for coding agents. Post-training shifted to reinforcement learning across thousands of environments and millions of sandboxed runs, and the model is available on Hugging Face. The source reports gains over Mellum2 on LiveCodeBench, AIME, GPQA Diamond, BFCL v4, IFEval, and SWE-bench Verified, and says it serves almost twice the tokens of Qwen3.5-9B under heavy load.

Why it matters: The post shows how reinforcement learning in real sandboxed environments changed a compact open model's repository work, with benchmark gains against Mellum2 and two peers.

AITencent Cloud has open-sourced Octop, a self-hosted multi-agent AI assistant platform aimed at families and small teams, with multi-user accounts and data kept on the user's own machine. The full text describes it as a single Python process that bundles the backend, web dashboard, CLI, IM gateway, cron jobs, and multi-agent runtime, with state rebuilt from SQLite on restart.

AIUniPat AI's PaperBenchX benchmark found the strongest model, GPT-6 Astra, fully reproduced only 13.98% of 93 real research-paper tasks across 12 scientific fields. Reproduction was judged by regenerating outputs in an isolated environment, with 3,168 expert-verified scoring items. UniPat has open-sourced 12 test tasks and kept 81 tasks closed to preserve long-term evaluation validity.

AIAnthropic has launched the Anthropic Cyber Mission, which starts with the Critical Infrastructure Defense Program for operational technology and OSS Scanner for open-source projects. The defense program brings frontier Claude models, on-site engineers and threat research to trusted providers such as Accenture, CrowdStrike and Palo Alto Networks. OSS Scanner gives enrolled open-source projects periodic free scans from its strongest models, with reports sent without human review and an expected true-positive rate above 90%.

Why it matters: The announcement shows how a frontier AI lab is packaging cyber defense around critical infrastructure and open-source maintainers, including the program's partners and access routes.

AIAnthropic is launching OSS Scanner, an opt-in service that runs periodic security scans of enrolled open-source projects using its strongest models at no cost. Its outputs are fully model-generated without human review, so some reports may be incorrect or invalid, though a pilot found 85 of 97 checked critical and high-severity findings met Anthropic's disclosure bar. Core maintainers of eligible projects can enroll through a GitHub pull request.

AIMicrosoft has repositioned Windows as the home for hybrid intelligence, where AI agents can run locally or in the cloud. The announcement covers four layers: MXC reaching general availability for agent isolation, local models such as MAI Code 1.1 Flash, Copilot on Copilot+ PCs gaining local context and actions in coming months, and new hardware including RTX Spark PCs and DGX Station for Windows.

AIThis Next Token episode discusses Personal Agents, including Dots in Codex, memory and cloud computer permissions, and whether agents should act as assistants or digital twins. The hosts also cover Instinct's booking and business-travel model, hands-on projects built with Opus 5.5, and whether software, games, and hardware could become open source as AI makes rewriting easier.

AIUnsloth Studio binds remote-code approval to a fingerprint of the scanned code, so changed code requires fresh consent before it runs. It also blocks weight files that Hugging Face has flagged for malware in the path the selected loader would deserialize. The article describes these checks as one layer among several, alongside package-content scans and OS sandboxes, and notes that the scanner is not a sandbox and cannot catch every evasion.

AIFrench AI company Mistral has released Mistral Large 4, its new flagship model with 1 trillion total parameters. The company plans to make all of the model's weights open at the end of the month.

AIEpoch gave six models 11 real work tasks from its own operations, including graphic design, data insights, and research design, and graded outputs against employee standards. Fable 5.1 and GPT-6 Astra led on average task performance, reliably handling well-defined work such as coding and computational analysis. The report finds that all models still fail on open-ended judgment, including matching Epoch's standards, designing informative experiments, and generating diverse ideas, so the authors conclude AI cannot yet replace workers at Epoch.

Why it matters: The report separates well-defined task reliability from open-ended judgment failures, which benchmark scores on easily verifiable tasks would miss.

AIA Hugging Face blog author used the ML-Intern agent in HuggingChat to build six small models by writing detailed prompts that specify datasets, base models, baselines, smoke tests, and spending limits. The projects include a citrus disease vision-language model, a Huggy character LoRA, a camera-angle LoRA, a doodle-to-object LoRA, a 0.8B prompt rewriter, and a 4-step distilled Agate model, with total compute cost of about USD 103. Each project's prompts and public models are linked from the post.

Why it matters: The author shows how prompt structure, baselines, smoke tests, and budget caps shape an agent-driven training workflow, with per-project costs given.

AIDeveloper Brandon Thomas's Artcraft has launched seven open source apps in Rust that aim to replicate the interfaces and tools of Adobe Photoshop, Illustrator, Premiere, Lightroom, After Effects, InDesign, and Acrobat Pro. Thomas said he used Anthropic's Claude Opus 5.5 to build the clean-room replacements, with WebAssembly versions available for browser use. The apps remain in a "super early alpha" state, and commenters have pointed out many current shortcomings.

AI…open-source inference with @IBM + @nvidia on a dedicated B300 cluster on IBM Cloud. Alan + our CRO Kai Mak 👇🏻

AI…returns typed answers with probabilities $0.02/M input. Output is free

AIThanks to @deepseek_ai for the model and kernels, @nvidia for the collaboration, and @SemiAnalysis_ for AgentX. Built by @inferact and the vLLM community.

AI…user on @SemiAnalysis_ AgentX. Here is how, with interactive figures you can step through 🧵

AIFrom the @Microsoft windows launch today with @satyanadella @sriramk @JensenHuang

AI…quickstarts. Quickstart: Docs:

AILocal sandboxing for GitHub Copilot is now generally available. It lets Copilot run commands in an isolated environment with controlled access to files, networks, system capabilities, and credentials. Enterprise teams can also centrally manage policies, and the feature is available in GitHub Copilot CLI, the GitHub Copilot app, and @code.

AIAi2 Comms Lead @Kyle_L_Wiggers catches up with Senior Director of NLP @nlpnoah to talk agentic models, the next iteration of Olmo, & more! At COLM? Come say hi at booth 302! 👋

AIThe software and hardware stacks are finally coming together. Our community has put a lot of hard work in the past years and it shows. Looking forward to more users embracing local AI.

AILiquid AI released two open-weight multimodal decision models, d1-3B and d1-omni-600M, which return probability answers in one forward pass with zero output tokens. d1-3B scores 48.57 on Decision Index v0.2.1 and answers one question in 8 ms on an RTX 4090, while the models are licensed free for commercial use below $10 million in annual revenue.

AISend text, JSON, or images and get back typed answers with probabilities instead of generated text $0.24/M input for Clef, $0.09/M for Clef Flash. Output is free

AI…we'll make it more accessible to regular users.

AId1-3B single-question latency, measured one request at a time: > NVIDIA RTX 4090: 8 ms > Jetson AGX Thor: 16 ms > Jetson AGX Orin 64 GB: 26 ms > Jetson Orin Nano: 50 ms 4/

AIBuilt from LFM2.5-VL-3B, it makes decisions from text and images in one pass. Use it for reranking, agent guardrails, and visual inspection. 2/

AI…investors including @nvidia @M12vc @SamsungNext @robotventures @usv @ycombinator @MenloVentures, among others

AIWe converted Qwen, Gemma, Llama all into decision models by fine-tuning using a Clef head, boosting accuracy from 30 to up to 78%. You can try it yourself via Unsloth Desktop!

AILlamaIndex announced OpenDocRouter, a single API that routes document parsing requests to any of 10 frontier and open-source models at launch, including Claude Opus 5.5, Gemini 3.8 Flash, GPT-6 Luna, MinerU2.5-Pro, and PaddleOCR-VL-1.6. Users can switch models in one line with the same request and markdown output, and each model is scored on ParseBench for quality and cost. Pricing is per-token, failed pages are not charged, and the service costs $0.86 to $48.82 per 1,000 pages depending on the model.

AIResearchers can build on these to test new architectures & train the full system without repeating that initial stage.