As of: 2026-10-10 14:45 CEST. Researched and written by AI agents (Claude Opus 5.5 in Claude Code). Human editor: Adam Bicz. Canonical page: https://postcutoff.com/news/open-source/ # AI news: Open source 46 events in Open source of 1,105 in the log, newest first. Page 1 of 1, 50 events per page, grouped by the day each event happened. ## Wednesday 7 October 2026 - ['Software is over'](https://postcutoff.com/e/2026-10-07-artcraft-opus-5-5-adobe-office-clones/) (Open source; ArtCraft, Anthropic, Adobe). It is the most visible test so far of the claim that frontier coding agents make large, established desktop software cheap to copy. Source: https://github.com/storytold/photocraft ## Tuesday 6 October 2026 - [Google DeepMind releases EmbeddingGemma 2](https://postcutoff.com/e/2026-10-06-embeddinggemma-2/) (Open source; Google DeepMind). A small, permissively licensed multimodal embedding model makes on-device multimodal search and RAG practical without a hosted API. Source: https://blog.google/innovation-and-ai/technology/developers-tools/embeddinggemma-2/ ## Sunday 4 October 2026 - [Axios: Nvidia-backed Reflection AI is about to release its first open-weight model, with other Western open-weight models due in October](https://postcutoff.com/e/2026-10-04-reflection-ai-first-open-weight-model-axios/) (Open source; Reflection AI, Nvidia). Chinese labs dominate open-weight usage (see the Vercel index entry). Source: https://reflection.ai/blog/introducing-beam ## Saturday 3 October 2026 - [Aleph Alpha releases Kolibri, a 78B-parameter (3.5B active) open-weight English–German 'sovereign' MoE model under Apache 2.0](https://postcutoff.com/e/2026-10-03-aleph-alpha-kolibri-open-weights/) (Open source; Aleph Alpha). Kolibri is a notable European open-weight release, from a lab that agreed in Sept 2026 to combine with Cohere. Source: https://aleph-alpha.com/en/blog/kolibri-has-landed-a-sovereign-open-weight-model/ ## Friday 2 October 2026 - [Nathan Lambert and Tom Zick launch Trillium Labs, a nonprofit for open post-training recipes and open frontier-AI science](https://postcutoff.com/e/2026-10-02-trillium-labs-launch/) (Open source; Trillium Labs). Fully open post-training recipes (as opposed to open weights alone) have been scarce. Source: https://x.com/natolambert/status/2106060179985019085 - [Ai2 open-sources AstaBrief, an 8B model (Qwen3-8B base) that writes cited scientific reports ~3.5x faster in Asta](https://postcutoff.com/e/2026-10-02-ai2-astabrief-open-report-model/) (Open source; Ai2). A minor release, but one of the few open-weight, purpose-built models for cited literature reports; it shows the small-model/fast-mode pattern spreading to research assistants. Source: https://allenai.org/blog/astabrief ## Thursday 1 October 2026 - [Cloudflare releases Clef and Clef-flash, open-weight multimodal decision models that are Jev-API compatible](https://postcutoff.com/e/2026-10-01-cloudflare-clef-decision-models/) (Open source; Cloudflare). It is the second big-company follower in two days, after AWS's Strands Decider 2B. Source: https://blog.cloudflare.com/clef-decision-models/ - [PewDiePie unveils Ajax, an 'uncensored' local 9B agent model, and says OpenAI banned him twice, once for distillation](https://postcutoff.com/e/2026-10-01-pewdiepie-ajax-openai-distillation-ban/) (Open source; PewDiePie, OpenAI). It brings the distillation fight between frontier labs and open-model builders, until now about Chinese labs, to a mass audience, and shows how detectable distillation-style use of APIs has become. Source: https://www.youtube.com/watch?v=ODDJXGY_1kQ ## Wednesday 30 September 2026 - [Bilibili open-sources Index-Translate, a 150-language translation model family up to 35B-A3B](https://postcutoff.com/e/2026-09-30-bilibili-index-translate/) (Open source; Bilibili). It is another open-weight Chinese model family built on Qwen. Source: https://github.com/bilibili/Index-Translate ## Tuesday 29 September 2026 - [Rest of World: ModelScope and MoArk compete as China's Hugging Face alternatives](https://postcutoff.com/e/2026-09-29-china-modelscope-moark-hugging-face-alternatives/) (Open source; Alibaba, OSChina). Chinese labs release weights on both Hugging Face and ModelScope. Source: https://restofworld.org/2026/china-open-source-ai-hugging-face-modelscope-moark/ ## Monday 21 September 2026 - [Xiaomi releases MiMo-V2.6 Pro (1.02T MoE) and Flash under MIT license](https://postcutoff.com/e/2026-09-21-xiaomi-mimo-v2-6/) (Open source; Xiaomi; major). The top open-weights model now comes from a consumer-electronics company rather than DeepSeek, Qwen or Moonshot, and it is MIT-licensed. Source: https://mimo.mi.com/models/en-US/mimo-v2.6-pro - [Tim Dettmers' dlab open-source week: 1.5-bit inference and a 125B model on one 24 GB GPU](https://postcutoff.com/e/2026-09-21-dettmers-dlab-open-source-week/) (Open source; dlab, Carnegie Mellon University). If the claims hold, 1.5-bit weights and single-consumer-GPU inference of 100B+ MoE models push open-weights AI further out of datacenters. Source: https://timdettmers.com/2026/09/21/dlab-open-source-week/ - [Yandex open-sources AliceAI-Foundation-80B-A3B-Base, an 80B/3B-active MoE trained from scratch (Apache 2.0)](https://postcutoff.com/e/2026-09-21-yandex-aliceai-foundation-80b-open-weights/) (Open source; Yandex). It is the most capable open-weights model from a Russian lab so far and adds a strong Russian-language base model under a permissive license. Source: https://ir.yandex/press-releases?year=2026&id=2026-09-21 ## Friday 18 September 2026 - [SAIR launches the Open Math Model initiative for community-governed open-weight math AI, plus Lean Kernel and Andrews–Curtis challenges](https://postcutoff.com/e/2026-09-18-sair-open-math-model-initiative/) (Open source; SAIR Foundation, Lean FRO, Caltech). It is the most concrete attempt by leading mathematicians to build an open, independent alternative to frontier labs' math AI, with governance and data-consent rules written in from the start. Source: https://terrytao.wordpress.com/2026/09/18/sairs-open-math-model-initiative/ ## Thursday 17 September 2026 - [PrismML releases Ternary Bonsai 2 27B](https://postcutoff.com/e/2026-09-17-prismml-ternary-bonsai-2-27b/) (Open source; PrismML). Near-lossless ternary compression at 27B scale makes capable reasoning models practical on local, low-power hardware. Source: https://prismml.com/news/bonsai-2-27b ## Thursday 10 September 2026 - [Unitree open-sources UnifoLM-WLA-1.0 humanoid foundation model](https://postcutoff.com/e/2026-09-10-unitree-unifolm-wla-1-0/) (Open source; Unitree Robotics). The world's highest-volume humanoid maker now ships an openly licensed foundation model for its own robots, lowering the barrier for G1 developers. Source: https://github.com/unitreerobotics/unifolm-wla ## Wednesday 9 September 2026 - [YuE2: open-weights song model that plans an editable score first, claims top WildSongBench score over Suno v5](https://postcutoff.com/e/2026-09-09-yue2-open-music-model/) (Open source; Multimodal Art Projection (M-A-P), HKUST). It is the strongest claim yet that an open model runnable on one consumer GPU matches the leading commercial song generator, released the same day Suno moved to licensed-data v6. Source: https://github.com/multimodal-art-projection/YuE ## Friday 28 August 2026 - [Tencent open-sources Hunyuan Hy4 preview (770B MoE, 1M+ context)](https://postcutoff.com/e/2026-08-28-tencent-hunyuan-hy4-preview/) (Open source; Tencent). Tencent joins DeepSeek, Moonshot, Alibaba and Zhipu in shipping ~1T-class open-weights models, deepening the Chinese open-model ecosystem. Source: https://pandaily.com/tencent-hunyuan-hy4-preview-open-source-aug2026 ## Wednesday 26 August 2026 - [Qwen3.8-Flash-Next, a 125B MoE with 6B active, previews the Qwen 4 architecture](https://postcutoff.com/e/2026-08-26-qwen3-8-flash-next/) (Open source; Alibaba, Qwen). It pushes the cost frontier: near-frontier agentic coding numbers at 6B active parameters make strong models cheap to serve at 1M-token contexts. Source: https://qwen.ai/blog?id=qwen3.8-flash-next ## Tuesday 25 August 2026 - [BreezeBlue releases Breeze TTS 2, the new top open-weights text-to-speech model](https://postcutoff.com/e/2026-08-25-breeze-tts-2/) (Open source; BreezeBlue). It pushed the open-weights ceiling in TTS about 90 Elo higher, narrowing the gap to closed leaders (Eleven v4, Cartesia Sonic-3.6). Source: https://huggingface.co/BreezeBlue/Breeze-TTS-2 ## Thursday 20 August 2026 - [Google's Gemma open models pass one billion downloads and 100,000+ community variants](https://postcutoff.com/e/2026-08-20-gemma-one-billion-downloads/) (Open source; Google DeepMind). On Aug 20, 2026 Google announced that its Gemma family of open-weight models had passed one billion cumulative downloads, with developers publishing more than 100,000 derived variants (the "Gemmaverse"). Source: https://blog.google/innovation-and-ai/technology/developers-tools/gemma-one-billion-downloads/ ## Thursday 13 August 2026 - [MiniMax open-sources Music 3.0, a five-minute full-song generator](https://postcutoff.com/e/2026-08-13-minimax-music-3-open-weights/) (Open source; MiniMax). Source: https://www.minimax.io/blog/minimax-music-3-0-next-generation-open-weights-production-ready-versatile-music-model ## Tuesday 11 August 2026 - [NVIDIA releases open Nemotron 3.5 Lightning and NeMo Switchyard model router](https://postcutoff.com/e/2026-08-11-nvidia-nemotron-3-5-lightning/) (Open source; NVIDIA). NVIDIA is now a significant American open-weight model developer; cheap local MoE workers plus routing directly target the cost of long-running agents, which dominate 2026 inference demand. Source: https://blogs.nvidia.com/blog/nemotron-lightning-switchyard-rtx-dgx/ ## Monday 10 August 2026 - [Meta returns to open weights with Muse Glimmer, a 30B Apache-2.0 agentic model](https://postcutoff.com/e/2026-08-10-meta-muse-glimmer-open-weights/) (Open source; Meta; major). On 2026-08-10 Meta released Muse Glimmer, a 30B-parameter open-weight model under Apache 2.0, optimized for local, always-on agent workflows and designed to run on a single consumer GPU or Mac. Source: https://research.meta.ai/blog/introducing-muse-glimmer-open-agentic-model ## Monday 3 August 2026 - [NVIDIA releases NemotronLabs VoiceChat 11B, an open full-duplex voice model with tool calling](https://postcutoff.com/e/2026-08-03-nvidia-nemotronlabs-voicechat/) (Open source; NVIDIA). Open full-duplex models (Kyutai Moshi, NVIDIA PersonaPlex) were mostly chat demos. Source: https://huggingface.co/nvidia/NVIDIA-NemotronLabs-VoiceChat-11B ## Thursday 16 July 2026 - [Xiaomi open-sources Xiaomi-Robotics-1, a VLA trained on 100K+ hours of real trajectories](https://postcutoff.com/e/2026-07-16-xiaomi-robotics-1/) (Open source; Xiaomi). It makes a 100K-hour-scale robot model openly available, narrowing the data gap between closed US labs and open Chinese releases. Source: https://arxiv.org/abs/2607.15330 ## Wednesday 15 July 2026 - [Thinking Machines Lab releases Inkling, its first open-weights model](https://postcutoff.com/e/2026-07-15-thinking-machines-inkling/) (Open source; Thinking Machines Lab; major). Source: https://thinkingmachines.ai/news/introducing-inkling/ ## Monday 13 July 2026 - [Xiaomi open-sources Xiaomi-Robotics-U0, a 38B unified world model that generates multi-view robot scenes and training data](https://postcutoff.com/e/2026-07-13-xiaomi-robotics-u0/) (Open source; Xiaomi). Like NVIDIA's Cosmos, it bets that generated data can offset the shortage of real robot data. Source: https://arxiv.org/abs/2607.11643 ## Thursday 2 July 2026 - [Mistral releases Leanstral 1.5, an open-weights Lean 4 proof agent that saturates miniF2F and solves 587/672 PutnamBench problems](https://postcutoff.com/e/2026-07-02-mistral-leanstral-1-5/) (Open source; Mistral AI). Formal proof was a frontier-lab capability (AlphaProof, Aristotle, Seed-Prover) through 2025. Source: https://mistral.ai/news/leanstral-1-5 ## Monday 1 June 2026 - [NVIDIA releases Cosmos 3, an open omni-model for physical AI](https://postcutoff.com/e/2026-06-01-nvidia-cosmos-3-open-release/) (Open source; NVIDIA). It is the largest open world model aimed at robotics and autonomous vehicles. Source: https://huggingface.co/blog/nvidia/cosmos-3-for-physical-ai ## Monday 23 March 2026 - [Mistral releases Voxtral TTS, an open-weight 4B text-to-speech model with 3-second voice cloning](https://postcutoff.com/e/2026-03-23-mistral-voxtral-tts/) (Open source; Mistral AI). With both open ASR and open TTS, Mistral became one of the few frontier labs offering a full open-weight voice stack, giving European and self-hosting customers an alternative to ElevenLabs and OpenAI voices. Source: https://mistral.ai/news/voxtral-tts ## Monday 9 March 2026 - [Fish Audio open-sources S2](https://postcutoff.com/e/2026-03-09-fish-audio-s2-open-source/) (Open source; Fish Audio). It made open TTS with fine-grained, LLM-style prompt control and production streaming available to the public, and set the open-weights bar for most of 2026. Source: https://fish.audio/blog/fish-audio-open-sources-s2/ ## Wednesday 28 January 2026 - [ACE-Step 1.5: MIT-licensed song generator that runs on consumer GPUs](https://postcutoff.com/e/2026-01-28-ace-step-1-5/) (Open source; ACE Studio, StepFun). It made near-commercial song generation practical on laptops and gaming GPUs under a permissive license, the open alternative to Suno/Udio in the year the commercial services moved to licensed data. Source: https://github.com/ace-step/ACE-Step-1.5 ## Thursday 22 January 2026 - [Alibaba open-sources Qwen3-TTS and, a week later, Qwen3-ASR](https://postcutoff.com/e/2026-01-22-qwen3-tts-asr-open-weights/) (Open source; Alibaba, Qwen). Voice cloning and voice design had mostly been proprietary (ElevenLabs and others). Source: https://github.com/QwenLM/Qwen3-TTS ## Thursday 27 November 2025 - [DeepSeekMath-V2: open-weights self-verifying prover reaches IMO 2025 gold level and 118/120 on Putnam 2024](https://postcutoff.com/e/2025-11-27-deepseekmath-v2/) (Open source; DeepSeek; major). It made olympiad-level natural-language proof generation reproducible outside the big US labs. Source: https://arxiv.org/abs/2511.22570 ## Tuesday 5 August 2025 - [OpenAI releases gpt-oss, its first open-weight LLMs since GPT-2](https://postcutoff.com/e/2025-08-05-gpt-oss/) (Open source; OpenAI). Gave the US a competitive open-weight reasoning model and ended OpenAI's six-year hiatus from open releases. Source: https://openai.com/index/introducing-gpt-oss/ ## Friday 11 July 2025 - [Moonshot AI releases Kimi K2, a 1-trillion-parameter open-weights agentic model](https://postcutoff.com/e/2025-07-11-kimi-k2/) (Open source; Moonshot AI). Part of a 2025 wave (DeepSeek, Qwen, Kimi, GLM) that made Chinese labs the leaders in open-weight models. Source: https://moonshotai.github.io/Kimi-K2/ ## Saturday 5 April 2025 - [Meta releases Llama 4 Scout and Maverick](https://postcutoff.com/e/2025-04-05-llama-4/) (Open source; Meta). Signaled Meta's loss of open-weights leadership to Chinese labs (DeepSeek, Qwen, Kimi) and precipitated its superintelligence reorganization. Source: https://ai.meta.com/blog/llama-4-multimodal-intelligence/ ## Monday 20 January 2025 - [DeepSeek-R1: open-weights reasoning model rivals o1 and shakes markets](https://postcutoff.com/e/2025-01-20-deepseek-r1/) (Open source; DeepSeek; historic). The 'DeepSeek moment' showed that frontier reasoning could be replicated cheaply and openly, triggering a market shock, a wave of open reasoning models, and US policy debates on China. Source: https://arxiv.org/abs/2501.12948 ## Thursday 26 December 2024 - [DeepSeek-V3: frontier-level open model trained for ~$5.6M in GPU time](https://postcutoff.com/e/2024-12-26-deepseek-v3/) (Open source; DeepSeek; historic). Upended assumptions about the cost of frontier AI and China's position; it was the base for DeepSeek-R1 weeks later. Source: https://arxiv.org/abs/2412.19437 ## Tuesday 23 July 2024 - [Llama 3.1 405B: the first frontier-class open-weights model](https://postcutoff.com/e/2024-07-23-llama-3-1-405b/) (Open source; Meta; major). Narrowed the open–closed gap to months; open frontier weights reshaped policy debates and enabled wide distillation. Source: https://ai.meta.com/blog/meta-llama-3-1/ ## Thursday 18 April 2024 - [Meta releases Llama 3 (8B, 70B)](https://postcutoff.com/e/2024-04-18-llama-3/) (Open source; Meta). Showed the benefit of training small models far past Chinchilla-optimal, and narrowed the open/closed gap. Source: https://ai.meta.com/blog/meta-llama-3/ ## Monday 11 December 2023 - [Mistral AI releases Mixtral 8x7B, an open mixture-of-experts model](https://postcutoff.com/e/2023-12-11-mixtral/) (Open source; Mistral AI). Popularized mixture-of-experts in open models (later used by DeepSeek-V3, Llama 4, Qwen) and established Europe's leading AI startup. Source: https://mistral.ai/news/mixtral-of-experts ## Tuesday 18 July 2023 - [Meta releases Llama 2 with a commercial-use license](https://postcutoff.com/e/2023-07-18-llama-2/) (Open source; Meta, Microsoft; major). Legitimized open-weights LLMs in industry and intensified the open vs. closed AI policy debate. Source: https://about.fb.com/news/2023/07/llama-2/ ## Friday 24 February 2023 - [Meta releases LLaMA, sparking the open-weights LLM wave](https://postcutoff.com/e/2023-02-24-llama/) (Open source; Meta AI; historic). Kick-started the open-weights LLM movement that later produced Llama 2/3, Mistral, Qwen and DeepSeek. Source: https://ai.meta.com/blog/large-language-model-llama-meta-ai/ ## Monday 22 August 2022 - [Stable Diffusion released as open weights](https://postcutoff.com/e/2022-08-22-stable-diffusion/) (Open source; Stability AI, CompVis (LMU Munich), Runway; historic). The open-weights release made generative AI a grassroots movement and set off legal battles over training data. Source: https://stability.ai/news/stable-diffusion-public-release Other views: All https://postcutoff.com/news/; Major only https://postcutoff.com/news/major/; Policy & safety https://postcutoff.com/news/policy-safety/; Science & math https://postcutoff.com/news/science/; Business https://postcutoff.com/news/business/; Model releases https://postcutoff.com/news/model-release/; Research https://postcutoff.com/news/research/; Chips & compute https://postcutoff.com/news/hardware-compute/; Products https://postcutoff.com/news/product/; Open source https://postcutoff.com/news/open-source/; Agents https://postcutoff.com/news/agents/; Robotics https://postcutoff.com/news/robotics/; Media generation https://postcutoff.com/news/media-generation/; Benchmarks https://postcutoff.com/news/benchmark/; Culture https://postcutoff.com/news/culture/; Milestones https://postcutoff.com/news/milestone/. Feeds: https://postcutoff.com/feeds/open-source.xml