#Open-source ecosystem

AIShanghai AI Lab's InternLM team released Intern-S2-397B, a multimodal foundation model for scientific intelligence and long-horizon agents. The model uses visual pre-training on raw scientific literature pages, multi-task reinforcement learning across more than 20 scientific domains, and agentic reinforcement learning in sandboxed environments.

AIMike Knoop says he sees a path to an ARC-AGI-4 benchmark focused on open-ended invention, which he calls the gating capability between zero-sum automation and positive-sum innovation. He argues that coordinated slowdown efforts would likely apply to everyone, including open-source work, and cites chain of thought and the transformer as inventions that grew out of open science research. He concludes the research frontier must stay open to keep humanity on a positive-sum path.

AIInferact says vLLM now supports DeepSeek v4.1 Flash on day zero across H100, H200, B200, B300, GB200, and GB300 GPUs. SemiAnalysis independently verified the NVIDIA support, while the post notes AMD vLLM still does not work with the model. Serving recipes are available at recipes.vllm.ai.

AINathan Lambert has compiled a reading list of open-model writing covering why labs release open weights, the open-versus-closed debate, and US-China competition, last updated 15 September 2026. The list includes pieces on open-model economics, safety and marginal-risk research, and recent Chinese releases such as Kimi K3 and GLM-5.2. It also cites lawmaker inquiries into Western companies' use of Chinese models.

AIShanghai Artificial Intelligence Laboratory has released Atria Dawn Preview, an agentic model built on the 744B-parameter MoE GLM-5.2 foundation model, with a 256K context window. The release page reports benchmark results across search, coding, tool use, productivity, and cybersecurity, and describes text-only setup for Codex and Claude Code.

Why it matters: The release page gives a full benchmark table against named rivals and setup steps for Codex and Claude Code, useful for anyone evaluating agentic models.

AITogether AI expanded its Fine-Tuning service with support for newer open-weight models, live metrics tracking, and finer training controls. Expert LoRA adapters can be applied to Mixture-of-Experts expert layers, and early stopping keeps the checkpoint with the best validation loss. Dataset previews, sample weights, pre-flight validation, and lower prices on selected models are also included.

AIAi2 released AstaBrief-8B-SFT, an 8B intermediate supervised fine-tuning checkpoint built on Qwen3-8B that turns a research question and retrieved literature excerpts into a cited report. On the ScholarQA-CS2 test set of 100 computer science questions, it scored an average of 83.7 versus 77.3 for base Qwen3-8B, with citation recall at 71.3 versus 64.6. The model is licensed under Apache 2.0 for research and educational use.

AIOllama announces that the DeepSeek V4.1 Flash model is available on its library, with a link to the model page. The post provides no further details about capabilities, parameters, or pricing.

AIGiles Thomas extended the GPT-2-style code from Sebastian Raschka's "Build a Large Language Model (from Scratch)" into a 6-expert, 2-active mixture-of-experts model and trained it from scratch over 8 days. Raschka praised the project as interesting LLM work done on a single GPU.

AICognition has welcomed Jonathan Kelley and the Dioxus team, whose framework Cognition used extensively to build and improve Devin's performance. Cognition plans to continue supporting Dioxus, Blitz, Taffy, and Subsecond while increasing investment in Dioxus-Native and Blitz. The Dioxus team will also work on Devin's virtual machine, computer use skills, and testing capabilities.

AIRadixArk links to a blog post on DeepSeek V4.1 at lmsys.org. The post text provided contains only the URL, so no further details on the model or its performance can be reported.

AIRadixArk has published a cookbook on its Miles documentation site covering how to run DeepSeek V4.1 Flash. The post itself contains only a link to the cookbook page, so no further details about features, figures, or setup steps are available.

AIRadixArk says Miles brings day-0 RL support to DeepSeek-V4.1-Flash, with SGLang providing inference support. The post says quantization-aware training mirrors SGLang's FP4/FP8 rounding, and that colocated training and rollout fit full-parameter RL on 16 GPUs. In a DAPO run over steps 0–80, per-token trainer–rollout KL stayed at 0.0012–0.0017 while reward rose from 0.51 to 0.78.

AISGLang and Miles ship day-0 inference and RL support for DeepSeek V4.1 Flash, with weights now available. The model is natively multimodal with 552B backbone parameters, 16B active during decode and 8B during prefill, and supports up to 1M context. V4.1 adds shared compressed KV across layers, a two-stage sparse indexer, and a 196B Engram lookup memory.

AIFireworks AI describes a four-stage path for teams moving from renting closed frontier models to training their own, starting with API use and prompt, context, and harness engineering. The post uses the UIPad computer-use dataset to show that Kimi K3 ties GPT 5.6 Sol overall at 87.7 but wins three of four categories while costing about half as much, suggesting routing. After roughly three hours of training on the training split, the tuned Kimi K3 outperforms GPT 5.6 Sol on the held-out test set.

AITinker says it and OpenResearch let agents test dozens of competing published post-training methods automatically, with compute cost forecast to within a dollar. The main post cites a grant-supported effort, while the quoted alphaXiv post says agents reproduced SDFT's continual learning benefits across Qwen3-8B and Qwen3-30B-A3B over multiple seeds.

AIRadixArk says its Miles framework natively supports SGLang for fast rollouts while keeping rollout and training aligned for reliable post-training at scale. The post thanks the community for contributions and feedback shaping Miles. A related post from @adarshxs describes Miles v0.1 running fully async agentic RL on a 744B MoE across 64 GB300 GPUs.

AIGoodfire used Ai2's open post-training stack, including the Dolci preference dataset, intermediate Olmo checkpoints, and OLMES evaluations, to trace a safety regression in Olmo. Preference training made Olmo more likely to comply with harmful requests on a refusal benchmark, and Goodfire linked part of this to specific Dolci examples where the preferred response encouraged compliance. Because Ai2 publishes the individual preferred and rejected responses, researchers could test targeted changes to reduce the regression.

AIGoogle announced general availability of ADK for Kotlin 1.0, a Kotlin Multiplatform framework for building AI agents on servers and Android. Version 1.0 reaches feature parity with ADK 1.0 Core and adds Android extensions for on-device models, cloud Gemini via Firebase AI Logic, and persistent sessions and memory with Room and AppSearch. The post includes a server-side incident triage example using KSP-generated tools and skills, plus an Android financial assistant example with human confirmation for transfers.

Why it matters: The post names the new Android and server-side capabilities and the code setup, helping Kotlin developers judge whether ADK fits their agent projects.

AIInferact says months of vLLM tuning for agentic workloads, validated on SemiAnalysis's AgentX benchmark, let open-source models reach up to 130K tokens per GPU-second. The company claims this is 106 times cheaper than Opus 5 API pricing. The work is described as part of a vLLM blog post covering architecture, framework, and runtime optimizations.

AICohere Labs has released Tiny Aya Base 32K, an open-weights pretrained model with 3.35 billion parameters and a 32K context window. The model covers 70+ languages, including many lower-resourced ones, and is designed for downstream adaptation and long-context research. It is a base model that has not been instruction-tuned, and it is licensed under CC-BY-NC.

AIUnsloth's Qwen3.8-27B GGUF is now the most-liked GGUF ever on Hugging Face, with the post predicting it will soon enter the top 30 most-liked models overall. Unsloth reports it reached 10M downloads and 3.7K likes in 24 days, and credits the community, Hugging Face, and the Qwen team.

AIOpen model licenses are tightening at the Chinese frontier, with Zhipu's GLM-5.3 switching from MIT to a custom license requiring a security review for inference and fine-tuning providers with over $10 billion in annual revenue. Motif-3 ships under an MIT license with strong scores for its size, while Tencent's Hy4-preview is a competent model that currently overthinks. Western makers Google and Meta have moved to Apache 2.0.

AIGoogle Antigravity announces that Google DeepMind's science-skills is available for download on GitHub. The post includes a link to the google-deepmind/science-skills repository and no further details about its contents or features.

AIMistral announced a €3 billion Series D round at a post-money valuation of more than €21 billion, led by Samsung Electronics with co-leads Scaleup Europe Fund and PSG Equity. The company says the funding will expand frontier research, compute capacity, infrastructure, and international growth, and that it now operates in 20 countries with 125+ enterprise customers including Airbus, ASML, and HSBC.

Why it matters: The round shows how a company frames sovereign, open-weight AI as a full stack spanning models, infrastructure, compute, and products, which is useful context for European enterprise AI strategy.

AINVIDIA's NV-Reason-CT is a 3D vision-language model for CT image analysis that combines a native 3D vision encoder with a language model. It is designed for radiology report generation, question answering, and multi-step reasoning across chest and abdominal CT volumes. The model converts a 384×384×384-mm input into 13,824 visual tokens without spatial downsampling and is available on Hugging Face under the OpenMDW-1.1 License.

AIMistral AI says its open-weight models, products, and infrastructure let organisations choose how and where they run AI, not just access a model. The company frames this as frontier performance without vendor lock-in.

AITencent Hunyuan says its Hy4 preview has been upgraded to reduce long thinking and over-verification on complex tasks, which users had flagged. The company reports the same task quality with fewer turns and lower input and output tokens, confirmed by benchmark and human evaluation. The upgrade is live for all users, and Tencent says it will keep iterating based on feedback.

AIOpenBMB released UltraData-Code-L2-Classifier, a suite of language-specific file-level scorers for 11 programming languages in UltraData-Code-L1. The L2 corpus selected with these scorers contains approximately 400B tokens and retains about 12.23% of L1 files, and a 10B-token test on a 1B model raised EvalPlus pass@1 by 7.80 points over L1 training.

AIOpenBMB has released MiniCPM5-2B, a dense 2B Transformer built for on-device and resource-constrained deployment, with an average score of 53.9 in its comparison set. The release also opens the UltraData datasets behind it, including UltraX, UltraData-Code, UltraData-SFT-Agent-2609 and UltraData-RL-2609, and includes GGUF, MLX, GPTQ and DSpark variants for common runtimes.

Why it matters: The release pairs a 2B model with open training datasets and reports per-benchmark comparisons against named same-size and larger models, letting readers check the claims directly.

AIA Berkeley Sky Lab researcher says stronger open models and new inference systems will make powerful local AI practical. The linked post reports Qwen3.8-Flash-Next running at 68.3 tok/s on a single RTX 5090 using an NVFP4 checkpoint, with 63GB host RAM and a 51GB n-gram table stored on NVMe at about 0.5% throughput cost.