{"apiVersion":"v1","taxonomyVersion":1,"generatedAt":"2026-10-29T13:28:15.274Z","data":{"id":"brf_iTi2ECDrYRcZ","kind":"weekly","slug":"2026-W39","window":{"from":"2026-09-21T00:00:00.000Z","to":"2026-09-28T00:00:00.000Z"},"version":1,"quiet":false,"title":"AI brief · 21 to 27 September 2026","headline":{"text":"New models and tools focus on typed decisions and long-context tasks","recordIds":["wsa_57ymh3i83893","wsa_nvbrm95963g4","wsa_zgkwinpjvy59","wsa_y3ctzu2k9fzt"]},"dek":{"text":"The week brought long-context and multimodal model releases, a cluster of typed-decision models and local tooling, and new retrieval, agent-building and creative applications.","recordIds":["wsa_57ymh3i83893","wsa_nvbrm95963g4","wsa_pji8rnnhx4ta","wsa_zgkwinpjvy59","wsa_y3ctzu2k9fzt","wsa_c2g526khda6e","wsa_wsnmvac57vvz","wsa_zf6wait4hxfw","wsa_i8vuawwn9f59","wsa_9hgqk3kqjkxn"]},"tldr":[{"text":"The week brought long-context and multimodal model releases, a cluster of typed-decision models and local tooling, and new retrieval, agent-building and creative applications.","recordIds":["wsa_57ymh3i83893","wsa_nvbrm95963g4","wsa_pji8rnnhx4ta","wsa_zgkwinpjvy59","wsa_y3ctzu2k9fzt","wsa_c2g526khda6e","wsa_wsnmvac57vvz","wsa_zf6wait4hxfw","wsa_i8vuawwn9f59","wsa_9hgqk3kqjkxn"]}],"stories":[{"heading":"Models","blocks":[{"kind":"paragraph","items":[{"text":"Xiaomi MiMo released MiMo-V2.6-Flash-MOPD, a multimodal sparse MoE model with a 1M-token context length and an upgrade aimed at mitigating repeated tool calls.","recordIds":["wsa_57ymh3i83893"]},{"text":"NaiveAI introduced Naive-N0.5-Flash, an open-weight 309B mixture-of-experts model with a native 1M-token context window, designed for coding and AI R&D.","recordIds":["wsa_nvbrm95963g4"]},{"text":"Several releases target typed decisions, including JevK5 v0.3, GLiNER2.5-multi-Decide, Open-Jev-27B-v1.1 and deepopen; they return or score choices, labels or probabilities rather than relying on free-form generation.","recordIds":["wsa_zgkwinpjvy59","wsa_gi33mhazhyx3","wsa_dw22epac2yz9","wsa_emd8f7pqef7i"]}]}]},{"heading":"Developer platforms","blocks":[{"kind":"paragraph","items":[{"text":"Ollaya offers local serving for decision models through a CLI, daemon, API and desktop app, while Rizzo Flow provides local typed decisions with probabilities through an API, CLI and browser playground.","recordIds":["wsa_y3ctzu2k9fzt","wsa_qjh56x9kdpqc"]},{"text":"Cohere introduced Compass Cloud, a managed private-beta enterprise search and retrieval service available through APIs, an MCP server and a Python SDK.","recordIds":["wsa_c2g526khda6e"]},{"text":"JevHarness helps create task-specific Jev harnesses, and Unreal Agent coordinates LLM turns, tools, persisted session history and asynchronous operations.","recordIds":["wsa_wsnmvac57vvz","wsa_w86nvqc4zfat"]}]}]},{"heading":"Applications","blocks":[{"kind":"paragraph","items":[{"text":"Runway brought its tools into DaVinci Resolve through a plugin available on paid plans.","recordIds":["wsa_zf6wait4hxfw"]},{"text":"Astra-Ares is an experimental CLI tool that adapts reasoning effort for selected GPT-6 models during Codex tasks; ZCode offers an AI coding workspace across desktop, browser and terminal-agent interfaces.","recordIds":["wsa_i8vuawwn9f59","wsa_9hgqk3kqjkxn"]}]}]}],"radar":[],"sections":[{"heading":"Models","category":"model","recordIds":["wsa_57ymh3i83893","wsa_nvbrm95963g4","wsa_pji8rnnhx4ta","wsa_54q3vbaapqkw","wsa_4zfvhvvzij2d","wsa_zgkwinpjvy59","wsa_spipwhu2mp47","wsa_gi33mhazhyx3","wsa_yd6924kdbwsq","wsa_q6ebbw4er26p","wsa_3u4cii4qvqcd","wsa_w99usrp5ri82","wsa_y7aqipj5jjqp","wsa_dw22epac2yz9","wsa_emd8f7pqef7i","wsa_ucyp4kzvdjcn"],"sentences":[{"text":"Xiaomi MiMo released MiMo-V2.6-Flash-MOPD, a multimodal sparse MoE model with a 1M-token context length and an upgrade aimed at mitigating repeated tool calls.","recordIds":["wsa_57ymh3i83893"]},{"text":"NaiveAI introduced Naive-N0.5-Flash, an open-weight 309B mixture-of-experts model with a native 1M-token context window, designed for coding and AI R&D.","recordIds":["wsa_nvbrm95963g4"]},{"text":"Several releases target typed decisions, including JevK5 v0.3, GLiNER2.5-multi-Decide, Open-Jev-27B-v1.1 and deepopen; they return or score choices, labels or probabilities rather than relying on free-form generation.","recordIds":["wsa_zgkwinpjvy59","wsa_gi33mhazhyx3","wsa_dw22epac2yz9","wsa_emd8f7pqef7i"]}]},{"heading":"Developer platforms","category":"developer platform","recordIds":["wsa_y3ctzu2k9fzt","wsa_c2g526khda6e","wsa_wsnmvac57vvz","wsa_qjh56x9kdpqc","wsa_525j9f47i4bg","wsa_w86nvqc4zfat"],"sentences":[{"text":"Ollaya offers local serving for decision models through a CLI, daemon, API and desktop app, while Rizzo Flow provides local typed decisions with probabilities through an API, CLI and browser playground.","recordIds":["wsa_y3ctzu2k9fzt","wsa_qjh56x9kdpqc"]},{"text":"Cohere introduced Compass Cloud, a managed private-beta enterprise search and retrieval service available through APIs, an MCP server and a Python SDK.","recordIds":["wsa_c2g526khda6e"]},{"text":"JevHarness helps create task-specific Jev harnesses, and Unreal Agent coordinates LLM turns, tools, persisted session history and asynchronous operations.","recordIds":["wsa_wsnmvac57vvz","wsa_w86nvqc4zfat"]}]},{"heading":"Applications","category":"application","recordIds":["wsa_zf6wait4hxfw","wsa_i8vuawwn9f59","wsa_9hgqk3kqjkxn"],"sentences":[{"text":"Runway brought its tools into DaVinci Resolve through a plugin available on paid plans.","recordIds":["wsa_zf6wait4hxfw"]},{"text":"Astra-Ares is an experimental CLI tool that adapts reasoning effort for selected GPT-6 models during Codex tasks; ZCode offers an AI coding workspace across desktop, browser and terminal-agent interfaces.","recordIds":["wsa_i8vuawwn9f59","wsa_9hgqk3kqjkxn"]}]}],"items":[{"recordId":"wsa_57ymh3i83893","versionId":"ver_Q6-vHYwkQxNx","eventType":"new_development","category":"model","title":"MiMo-V2.6-Flash-MOPD","maker":"Xiaomi MiMo","summary":"MiMo-V2.6-Flash-MOPD is a multimodal sparse MoE model with 309B total parameters and 15B activated parameters. It has a 1M-token context length and is an MOPD upgrade of the MiMo-V2.6-Flash-RL checkpoint that mitigates repeated tool calls.","releaseNote":null,"classifications":{"developmentType":"model","capability":["text generation","analysis","agents"],"modality":["text","image","video","audio","multimodal"],"access":["open weights","API"],"deployment":["hosted","on-premises"],"pricing":["unknown"],"targetUser":["developer","researcher","enterprise"]},"citations":[{"url":"https://huggingface.co/XiaomiMiMo/MiMo-V2.6-Flash-MOPD","authority":"owner-controlled","publisher":"Hugging Face"}],"score":1,"confidence":"high","correctionState":"current","publishedAt":"2026-09-27T04:01:00.000Z","effectiveAt":"2026-09-27T20:05:24.181Z","url":"https://wowsuchai.withglint.app/records/wsa_57ymh3i83893","apiUrl":"https://wowsuchai.withglint.app/api/v1/records/wsa_57ymh3i83893","webUrl":"https://wowsuchai.withglint.app/records/wsa_57ymh3i83893"},{"recordId":"wsa_nvbrm95963g4","versionId":"ver_ZDF5jd230Ocf","eventType":"new_development","category":"model","title":"Naive-N0.5-Flash","maker":"NaiveAI","summary":"Naive-N0.5-Flash is an open-weight 309B mixture-of-experts model with 15.5B active parameters, designed for coding and AI R&D. It has a native 1M-token context window and uses hybrid Sliding-Window Attention and DeepSeek Sparse Attention.","releaseNote":null,"classifications":{"developmentType":"model","capability":["text generation","code generation","analysis"],"modality":["text","code"],"access":["open weights"],"deployment":["on-premises"],"pricing":["usage-based"],"targetUser":["developer","researcher"]},"citations":[{"url":"https://huggingface.co/NaiveAI/Naive-N0.5-Flash","authority":"owner-controlled","publisher":"Hugging Face"}],"score":1,"confidence":"high","correctionState":"current","publishedAt":"2026-09-27T14:14:32.000Z","effectiveAt":"2026-09-27T20:05:22.061Z","url":"https://wowsuchai.withglint.app/records/wsa_nvbrm95963g4","apiUrl":"https://wowsuchai.withglint.app/api/v1/records/wsa_nvbrm95963g4","webUrl":"https://wowsuchai.withglint.app/records/wsa_nvbrm95963g4"},{"recordId":"wsa_pji8rnnhx4ta","versionId":"ver_u4Mtuu0YWHZI","eventType":"new_development","category":"model","title":"MiniMax-H3 T2VA Veda Predictor 8 NFE 600-Step Preview","maker":"Veda-Sparse","summary":"This preview checkpoint is a tile-score predictor for MiniMax-H3 text-to-audio-video generation at eight denoising steps. It selects attention tiles for block-sparse attention at a 10% keep ratio, with twelve geometry plans packed into the weights file.","releaseNote":null,"classifications":{"developmentType":"model","capability":["analysis"],"modality":["text","audio","video","multimodal"],"access":["open weights","CLI"],"deployment":["on-premises"],"pricing":["unknown"],"targetUser":["developer"]},"citations":[{"url":"https://huggingface.co/Veda-Sparse/Minimax-H3-T2VA-Veda-8NFE-600Step-Preview","authority":"owner-controlled","publisher":"Hugging Face"}],"score":1,"confidence":"high","correctionState":"current","publishedAt":"2026-09-25T10:56:07.000Z","effectiveAt":"2026-09-26T19:59:22.179Z","url":"https://wowsuchai.withglint.app/records/wsa_pji8rnnhx4ta","apiUrl":"https://wowsuchai.withglint.app/api/v1/records/wsa_pji8rnnhx4ta","webUrl":"https://wowsuchai.withglint.app/records/wsa_pji8rnnhx4ta"},{"recordId":"wsa_54q3vbaapqkw","versionId":"ver_91X0E8YLxKiW","eventType":"new_development","category":"model","title":"Nanosaur2-670M","maker":"well9472","summary":"Nanosaur2-670M is a 670-million-parameter diffusion transformer for text-to-image illustration generation. It is released with ComfyUI support and model files for local inference.","releaseNote":null,"classifications":{"developmentType":"model","capability":["image generation","fine-tuning"],"modality":["text","image"],"access":["open weights"],"deployment":["on-premises"],"pricing":["unknown"],"targetUser":["creator","developer","researcher"]},"citations":[{"url":"https://huggingface.co/well9472/Nanosaur2-670M","authority":"owner-controlled","publisher":"Hugging Face"}],"score":1,"confidence":"high","correctionState":"current","publishedAt":"2026-09-25T14:54:41.000Z","effectiveAt":"2026-09-26T07:56:21.250Z","url":"https://wowsuchai.withglint.app/records/wsa_54q3vbaapqkw","apiUrl":"https://wowsuchai.withglint.app/api/v1/records/wsa_54q3vbaapqkw","webUrl":"https://wowsuchai.withglint.app/records/wsa_54q3vbaapqkw"},{"recordId":"wsa_4zfvhvvzij2d","versionId":"ver_Yqdulla5CehF","eventType":"new_development","category":"model","title":"Agate Preview 001","maker":"LogoLabs","summary":"Agate Preview 001 is a 260-million-parameter text-to-image model that generates 256 × 256 images from English prompts. Its weights are available under the MIT licence, and the release is identified as a research preview.","releaseNote":null,"classifications":{"developmentType":"model","capability":["image generation"],"modality":["text","image"],"access":["API","open weights","web"],"deployment":["on-device"],"pricing":["free","open-source"],"targetUser":["developer","researcher","creator"]},"citations":[{"url":"https://huggingface.co/Logolabs/agate-preview-001","authority":"owner-controlled","publisher":"Hugging Face"}],"score":1,"confidence":"high","correctionState":"current","publishedAt":"2026-09-25T14:11:48.000Z","effectiveAt":"2026-09-26T01:54:27.585Z","url":"https://wowsuchai.withglint.app/records/wsa_4zfvhvvzij2d","apiUrl":"https://wowsuchai.withglint.app/api/v1/records/wsa_4zfvhvvzij2d","webUrl":"https://wowsuchai.withglint.app/records/wsa_4zfvhvvzij2d"},{"recordId":"wsa_zgkwinpjvy59","versionId":"ver_cfwH0DXxK-PQ","eventType":"new_development","category":"model","title":"JevK5 v0.3","maker":"alibiserikbay","summary":"JevK5 v0.3 is an open-weight model based on Qwen3.5-4B, fine-tuned for typed decisions. It returns probabilities for options in one forward pass without generating tokens.","releaseNote":null,"classifications":{"developmentType":"model","capability":["analysis"],"modality":["text"],"access":["open weights","API"],"deployment":["on-premises"],"pricing":["unknown"],"targetUser":["developer","researcher","enterprise"]},"citations":[{"url":"https://huggingface.co/alibiserikbay/JevK5","authority":"owner-controlled","publisher":"Hugging Face"}],"score":1,"confidence":"high","correctionState":"current","publishedAt":"2026-09-22T13:53:01.000Z","effectiveAt":"2026-09-25T07:48:46.633Z","url":"https://wowsuchai.withglint.app/records/wsa_zgkwinpjvy59","apiUrl":"https://wowsuchai.withglint.app/api/v1/records/wsa_zgkwinpjvy59","webUrl":"https://wowsuchai.withglint.app/records/wsa_zgkwinpjvy59"},{"recordId":"wsa_spipwhu2mp47","versionId":"ver_DePB81TuAC-n","eventType":"new_development","category":"model","title":"Swift 1.5 Qwen3.8-Flash-Next","maker":"UkisAI","summary":"Swift 1.5 Qwen3.8-Flash-Next is UkisAI’s reasoning-efficient derivative of Qwen3.8-Flash-Next, released as a BF16 checkpoint with quantized variants. UkisAI reports 63.4% fewer median thinking tokens and 1.8× speedup at xhigh, with less than 1% accuracy loss compared with the base model.","releaseNote":null,"classifications":{"developmentType":"model","capability":["analysis","code generation","agents"],"modality":["text","code"],"access":["API","open weights"],"deployment":["hosted","on-premises"],"pricing":["free","enterprise","unknown"],"targetUser":["developer","researcher"]},"citations":[{"url":"https://huggingface.co/ukisai/Swift-Qwen3.8-Flash-Next","authority":"owner-controlled","publisher":"Hugging Face"}],"score":1,"confidence":"high","correctionState":"current","publishedAt":"2026-09-22T08:54:04.000Z","effectiveAt":"2026-09-25T07:48:44.475Z","url":"https://wowsuchai.withglint.app/records/wsa_spipwhu2mp47","apiUrl":"https://wowsuchai.withglint.app/api/v1/records/wsa_spipwhu2mp47","webUrl":"https://wowsuchai.withglint.app/records/wsa_spipwhu2mp47"},{"recordId":"wsa_gi33mhazhyx3","versionId":"ver_yg96NQPRKzPt","eventType":"new_development","category":"model","title":"GLiNER2.5-multi-Decide","maker":"Fastino","summary":"GLiNER2.5-multi-Decide is a 287M-parameter multilingual classification model in the GLiNER2.5 family. It accepts caller-provided labels and can score multiple classification heads in one forward pass.","releaseNote":null,"classifications":{"developmentType":"model","capability":["analysis"],"modality":["text"],"access":["open weights","SDK"],"deployment":["on-premises"],"pricing":["unknown"],"targetUser":["developer","enterprise"]},"citations":[{"url":"https://huggingface.co/fastino/GLiNER2.5-multi-Decide","authority":"owner-controlled","publisher":"Hugging Face"}],"score":1,"confidence":"high","correctionState":"current","publishedAt":"2026-09-24T15:32:46.000Z","effectiveAt":"2026-09-25T07:48:42.523Z","url":"https://wowsuchai.withglint.app/records/wsa_gi33mhazhyx3","apiUrl":"https://wowsuchai.withglint.app/api/v1/records/wsa_gi33mhazhyx3","webUrl":"https://wowsuchai.withglint.app/records/wsa_gi33mhazhyx3"},{"recordId":"wsa_yd6924kdbwsq","versionId":"ver_a7jD4fzYnXTE","eventType":"new_development","category":"model","title":"LFM2.5-VL-DSpark","maker":"Liquid AI","summary":"LFM2.5-VL-DSpark is an experimental vision-language draft model for speculative decoding with LFM2.5-VL-3B. Liquid AI reports decoding throughput improvements of up to 2.66× on GPUs and 3.13× on edge devices, with output quality unchanged.","releaseNote":null,"classifications":{"developmentType":"model","capability":["text generation"],"modality":["text","image","multimodal"],"access":["open weights"],"deployment":["on-device","on-premises"],"pricing":["unknown"],"targetUser":["developer","researcher"]},"citations":[{"url":"https://www.liquid.ai/blog/lfm2-5-vl-dspark","authority":"owner-controlled","publisher":"Liquid AI"},{"url":"https://huggingface.co/blog/LiquidAI/lfm2-5-vl-dspark","authority":"other","publisher":"Hugging Face"}],"score":1,"confidence":"high","correctionState":"current","publishedAt":"2026-09-24T00:00:00.000Z","effectiveAt":"2026-09-24T19:44:43.636Z","url":"https://wowsuchai.withglint.app/records/wsa_yd6924kdbwsq","apiUrl":"https://wowsuchai.withglint.app/api/v1/records/wsa_yd6924kdbwsq","webUrl":"https://wowsuchai.withglint.app/records/wsa_yd6924kdbwsq"},{"recordId":"wsa_q6ebbw4er26p","versionId":"ver_Xvm0VjSZGgmF","eventType":"new_development","category":"model","title":"Bespoke-Nimble-9B","maker":"Bespoke Labs","summary":"Bespoke-Nimble-9B is a Qwen3.5-9B LoRA adapter for context-grounded choices, booleans, and rubric score levels. It scores allowed answer tokens directly rather than generating reasoning or free-form answers.","releaseNote":null,"classifications":{"developmentType":"model","capability":["extraction","analysis"],"modality":["text"],"access":["open weights"],"deployment":["on-premises"],"pricing":["unknown"],"targetUser":["developer","researcher"]},"citations":[{"url":"https://huggingface.co/bespokelabs/Bespoke-Nimble-9B","authority":"owner-controlled","publisher":"Bespoke Labs"}],"score":1,"confidence":"high","correctionState":"current","publishedAt":"2026-09-18T01:18:02.000Z","effectiveAt":"2026-09-24T13:42:08.429Z","url":"https://wowsuchai.withglint.app/records/wsa_q6ebbw4er26p","apiUrl":"https://wowsuchai.withglint.app/api/v1/records/wsa_q6ebbw4er26p","webUrl":"https://wowsuchai.withglint.app/records/wsa_q6ebbw4er26p"},{"recordId":"wsa_3u4cii4qvqcd","versionId":"ver_w_mUMRtRFGej","eventType":"new_development","category":"model","title":"Meridian","maker":"Viggle","summary":"Meridian is a geometry-guided video model released as two LoRA adapters for the MiniMax-H3 transformer. It generates new camera viewpoints of existing video, and can also create camera movement from a single image.","releaseNote":null,"classifications":{"developmentType":"model","capability":["video generation"],"modality":["video","image"],"access":["CLI","open weights"],"deployment":["on-premises"],"pricing":["unknown"],"targetUser":["creator","developer","researcher"]},"citations":[{"url":"https://huggingface.co/Viggle/Meridian","authority":"owner-controlled","publisher":"Viggle"}],"score":1,"confidence":"high","correctionState":"current","publishedAt":"2026-09-14T15:07:00.000Z","effectiveAt":"2026-09-24T13:42:00.750Z","url":"https://wowsuchai.withglint.app/records/wsa_3u4cii4qvqcd","apiUrl":"https://wowsuchai.withglint.app/api/v1/records/wsa_3u4cii4qvqcd","webUrl":"https://wowsuchai.withglint.app/records/wsa_3u4cii4qvqcd"},{"recordId":"wsa_w99usrp5ri82","versionId":"ver_n95GSMECyS6Q","eventType":"new_development","category":"model","title":"Qwen-Image-2.1-viggle-turbo v0.2.1","maker":"Viggle","summary":"Qwen-Image-2.1-viggle-turbo v0.2.1 is a LoRA adapter distilled from Qwen-Image-2.1 for text-to-image generation and instruction-driven image editing. It supports editing with one to three reference images and is sampled in six transformer passes without classifier-free guidance.","releaseNote":null,"classifications":{"developmentType":"model","capability":["image generation"],"modality":["text","image"],"access":["open weights"],"deployment":["on-premises"],"pricing":["unknown"],"targetUser":["developer","creator"]},"citations":[{"url":"https://huggingface.co/Viggle/Qwen-Image-2.1-viggle-turbo","authority":"owner-controlled","publisher":"Viggle"}],"score":1,"confidence":"high","correctionState":"current","publishedAt":"2026-09-22T04:14:56.000Z","effectiveAt":"2026-09-24T13:41:59.302Z","url":"https://wowsuchai.withglint.app/records/wsa_w99usrp5ri82","apiUrl":"https://wowsuchai.withglint.app/api/v1/records/wsa_w99usrp5ri82","webUrl":"https://wowsuchai.withglint.app/records/wsa_w99usrp5ri82"},{"recordId":"wsa_y7aqipj5jjqp","versionId":"ver_Kmh4gG9aCA4B","eventType":"new_development","category":"model","title":"Laya Multilingual Core ML ANE","maker":"mizorewww","summary":"Laya Multilingual Core ML ANE is a downloadable FP16 checkpoint for short decisions, listed with CPU and Neural Engine support. The bundle has a B1/L96 capacity and a 96-token total limit.","releaseNote":null,"classifications":{"developmentType":"model","capability":["analysis","extraction"],"modality":["text"],"access":["API","SDK","CLI","open weights"],"deployment":["on-device"],"pricing":["open-source","free"],"targetUser":["developer"]},"citations":[{"url":"https://github.com/mizorewww/laya-coreml","authority":"owner-controlled","publisher":"GitHub new repositories"}],"score":1,"confidence":"high","correctionState":"current","publishedAt":"2026-09-19T23:14:13.000Z","effectiveAt":"2026-09-24T13:41:53.407Z","url":"https://wowsuchai.withglint.app/records/wsa_y7aqipj5jjqp","apiUrl":"https://wowsuchai.withglint.app/api/v1/records/wsa_y7aqipj5jjqp","webUrl":"https://wowsuchai.withglint.app/records/wsa_y7aqipj5jjqp"},{"recordId":"wsa_dw22epac2yz9","versionId":"ver_OQD0g83NSo6d","eventType":"new_development","category":"model","title":"Open-Jev-27B-v1.1","maker":"Zefan Cai","summary":"Open-Jev-27B-v1.1 produces typed probabilities for decisions from a supplied context, questions, and candidates, without autoregressive answer generation or generated-JSON parsing. Its released package contains LoRA adapters, a trained scalar decision head, and a saved calibration temperature; it requires pinned upstream Qwen weights and the Open-Jev loader.","releaseNote":null,"classifications":{"developmentType":"model","capability":["analysis","extraction"],"modality":["text"],"access":["open weights","CLI"],"deployment":["on-premises"],"pricing":["unknown"],"targetUser":["developer"]},"citations":[{"url":"https://github.com/Zefan-Cai/Open-Jev","authority":"owner-controlled","publisher":"GitHub new repositories"}],"score":1,"confidence":"high","correctionState":"current","publishedAt":"2026-09-20T20:12:13.000Z","effectiveAt":"2026-09-24T13:41:41.356Z","url":"https://wowsuchai.withglint.app/records/wsa_dw22epac2yz9","apiUrl":"https://wowsuchai.withglint.app/api/v1/records/wsa_dw22epac2yz9","webUrl":"https://wowsuchai.withglint.app/records/wsa_dw22epac2yz9"},{"recordId":"wsa_emd8f7pqef7i","versionId":"ver_Un-1Q6KIF4-N","eventType":"new_development","category":"model","title":"deepopen","maker":"DeepOpen","summary":"deepopen is an open-weight, non-autoregressive model for typed decisions, based on ModernBERT-large with 421 million parameters. It supports a 512-token context and is intended for English-language tasks.","releaseNote":null,"classifications":{"developmentType":"model","capability":["analysis","extraction","automation"],"modality":["text"],"access":["API","open weights"],"deployment":["on-premises"],"pricing":["open-source","free"],"targetUser":["developer","enterprise"]},"citations":[{"url":"https://github.com/deepopen-com/deepopen","authority":"owner-controlled","publisher":"GitHub new repositories"}],"score":1,"confidence":"high","correctionState":"current","publishedAt":"2026-09-21T02:12:24.000Z","effectiveAt":"2026-09-24T13:41:30.569Z","url":"https://wowsuchai.withglint.app/records/wsa_emd8f7pqef7i","apiUrl":"https://wowsuchai.withglint.app/api/v1/records/wsa_emd8f7pqef7i","webUrl":"https://wowsuchai.withglint.app/records/wsa_emd8f7pqef7i"},{"recordId":"wsa_ucyp4kzvdjcn","versionId":"ver_xZL0jtBfOzcZ","eventType":"new_development","category":"model","title":"AgentJev-0.6B","maker":"aimeigaoshou","summary":"AgentJev-0.6B is a coding-completion checkpoint that returns a probability distribution about whether an implementation meets stated requirements. It uses a Qwen3-0.6B base and was trained with supervised training and RLCD on executed coding pairs.","releaseNote":null,"classifications":{"developmentType":"model","capability":["analysis","agents"],"modality":["text"],"access":["open weights","API"],"deployment":["on-premises"],"pricing":["unknown"],"targetUser":["developer"]},"citations":[{"url":"https://huggingface.co/aimeigaoshou/agent-jev","authority":"owner-controlled","publisher":"Hugging Face"},{"url":"https://github.com/malevrigns/agent-jev","authority":"owner-controlled","publisher":"GitHub new repositories"}],"score":1,"confidence":"high","correctionState":"current","publishedAt":"2026-09-21T17:03:51.000Z","effectiveAt":"2026-09-24T13:41:17.069Z","url":"https://wowsuchai.withglint.app/records/wsa_ucyp4kzvdjcn","apiUrl":"https://wowsuchai.withglint.app/api/v1/records/wsa_ucyp4kzvdjcn","webUrl":"https://wowsuchai.withglint.app/records/wsa_ucyp4kzvdjcn"},{"recordId":"wsa_y3ctzu2k9fzt","versionId":"ver_m-_ESmJTHFwr","eventType":"new_development","category":"developer platform","title":"Ollaya","maker":"ollaya-dev","summary":"Ollaya is an open-source local serving platform for decision models, with a CLI, daemon, API, and desktop app. It pulls models by name and returns calibrated decisions from typed questions without generating text.","releaseNote":null,"classifications":{"developmentType":"developer platform","capability":["analysis","extraction","agents"],"modality":["text","multimodal"],"access":["API","SDK","CLI","desktop","open weights"],"deployment":["on-device","on-premises"],"pricing":["free","open-source"],"targetUser":["developer"]},"citations":[{"url":"https://github.com/ollaya-dev/ollaya","authority":"owner-controlled","publisher":"GitHub new repositories"}],"score":1,"confidence":"high","correctionState":"current","publishedAt":"2026-09-23T22:11:15.000Z","effectiveAt":"2026-09-26T07:56:23.175Z","url":"https://wowsuchai.withglint.app/records/wsa_y3ctzu2k9fzt","apiUrl":"https://wowsuchai.withglint.app/api/v1/records/wsa_y3ctzu2k9fzt","webUrl":"https://wowsuchai.withglint.app/records/wsa_y3ctzu2k9fzt"},{"recordId":"wsa_c2g526khda6e","versionId":"ver_s9vL94ZyZSqc","eventType":"new_development","category":"developer platform","title":"Compass Cloud","maker":"Cohere","summary":"Compass Cloud is Cohere’s managed private-beta offering for enterprise search and retrieval. It provides a configurable retrieval service through APIs, an MCP server, and a Python SDK, while Cohere manages the pipeline and model inference.","releaseNote":null,"classifications":{"developmentType":"developer platform","capability":["retrieval","search","reranking","agents"],"modality":["text","image","multimodal"],"access":["API","SDK"],"deployment":["hosted"],"pricing":["unknown"],"targetUser":["developer","enterprise"]},"citations":[{"url":"https://cohere.com/blog/compass-cloud-beta","authority":"owner-controlled","publisher":"Cohere"}],"score":1,"confidence":"high","correctionState":"current","publishedAt":"2026-09-25T13:10:00.000Z","effectiveAt":"2026-09-25T19:52:26.676Z","url":"https://wowsuchai.withglint.app/records/wsa_c2g526khda6e","apiUrl":"https://wowsuchai.withglint.app/api/v1/records/wsa_c2g526khda6e","webUrl":"https://wowsuchai.withglint.app/records/wsa_c2g526khda6e"},{"recordId":"wsa_wsnmvac57vvz","versionId":"ver_3cukHJ8Ab09I","eventType":"new_development","category":"developer platform","title":"JevHarness","maker":"TianyuCodings","summary":"JevHarness is a Python project and agent skill for creating task-specific Jev harnesses with an authoring LLM. It can optionally use rewards and complete execution traces to improve a harness, then freeze the selected strategy so runtime decisions use the harness code and Jev calls without an authoring LLM on every decision.","releaseNote":null,"classifications":{"developmentType":"developer platform","capability":["agents","automation"],"modality":["text","code"],"access":["CLI","SDK"],"deployment":["hosted"],"pricing":["unknown"],"targetUser":["developer"]},"citations":[{"url":"https://github.com/TianyuCodings/JevHarness","authority":"owner-controlled","publisher":"GitHub new repositories"}],"score":1,"confidence":"high","correctionState":"current","publishedAt":"2026-09-21T08:59:46.000Z","effectiveAt":"2026-09-24T19:44:55.878Z","url":"https://wowsuchai.withglint.app/records/wsa_wsnmvac57vvz","apiUrl":"https://wowsuchai.withglint.app/api/v1/records/wsa_wsnmvac57vvz","webUrl":"https://wowsuchai.withglint.app/records/wsa_wsnmvac57vvz"},{"recordId":"wsa_qjh56x9kdpqc","versionId":"ver_2nPIazQ5A5-1","eventType":"new_development","category":"developer platform","title":"Rizzo Flow","maker":"Rizzo AI Academy","summary":"Rizzo Flow is an open-source, local-first tool that turns unstructured state into typed decisions with probabilities using open-weight models. It provides a local API, command-line interface, and browser playground, and generates no output tokens for its decisions.","releaseNote":null,"classifications":{"developmentType":"developer platform","capability":["analysis","automation"],"modality":["text"],"access":["CLI","API","open weights"],"deployment":["on-premises","on-device"],"pricing":["open-source","unknown"],"targetUser":["developer"]},"citations":[{"url":"https://github.com/Rizzo-AI-Academy/rizzo-flow","authority":"owner-controlled","publisher":"GitHub new repositories"}],"score":1,"confidence":"high","correctionState":"current","publishedAt":"2026-09-21T00:00:02.000Z","effectiveAt":"2026-09-24T13:41:37.551Z","url":"https://wowsuchai.withglint.app/records/wsa_qjh56x9kdpqc","apiUrl":"https://wowsuchai.withglint.app/api/v1/records/wsa_qjh56x9kdpqc","webUrl":"https://wowsuchai.withglint.app/records/wsa_qjh56x9kdpqc"},{"recordId":"wsa_525j9f47i4bg","versionId":"ver_zdkB5fNUDIto","eventType":"new_development","category":"developer platform","title":"Laya-CoreML","maker":"mizorewww","summary":"Laya-CoreML is an independent Python and CLI package for running Laya typed-decision models with Apple Core ML. It supports local inference on Apple Silicon and can use downloadable model bundles, including Neural Engine variants.","releaseNote":null,"classifications":{"developmentType":"developer platform","capability":["analysis","extraction"],"modality":["text"],"access":["API","SDK","CLI","open weights"],"deployment":["on-device"],"pricing":["open-source","free"],"targetUser":["developer"]},"citations":[{"url":"https://github.com/mizorewww/laya-coreml","authority":"owner-controlled","publisher":"GitHub new repositories"}],"score":1,"confidence":"high","correctionState":"current","publishedAt":"2026-09-19T23:14:13.000Z","effectiveAt":"2026-09-24T13:41:36.203Z","url":"https://wowsuchai.withglint.app/records/wsa_525j9f47i4bg","apiUrl":"https://wowsuchai.withglint.app/api/v1/records/wsa_525j9f47i4bg","webUrl":"https://wowsuchai.withglint.app/records/wsa_525j9f47i4bg"},{"recordId":"wsa_w86nvqc4zfat","versionId":"ver_YRhqrrZV3NBx","eventType":"new_development","category":"developer platform","title":"Unreal Agent","maker":"Unreal Labs","summary":"Unreal Agent is an async-first agent harness with a Go library and executables. It coordinates LLM turns, tools, persisted session history, and asynchronous operations.","releaseNote":null,"classifications":{"developmentType":"developer platform","capability":["agents","automation"],"modality":["text","code"],"access":["SDK"],"deployment":["on-premises"],"pricing":["unknown"],"targetUser":["developer"]},"citations":[{"url":"https://github.com/unreallabsai/unreal-agent","authority":"owner-controlled","publisher":"GitHub new repositories"}],"score":1,"confidence":"high","correctionState":"current","publishedAt":"2026-09-21T23:14:02.000Z","effectiveAt":"2026-09-24T13:41:22.956Z","url":"https://wowsuchai.withglint.app/records/wsa_w86nvqc4zfat","apiUrl":"https://wowsuchai.withglint.app/api/v1/records/wsa_w86nvqc4zfat","webUrl":"https://wowsuchai.withglint.app/records/wsa_w86nvqc4zfat"},{"recordId":"wsa_zf6wait4hxfw","versionId":"ver_ukgRCEGs34_3","eventType":"new_development","category":"application","title":"Runway DaVinci Resolve Plugin","maker":"Runway","summary":"The Runway DaVinci Resolve Plugin brings Runway tools into DaVinci Resolve. It is available on paid plans.","releaseNote":null,"classifications":{"developmentType":"application","capability":["video generation"],"modality":["video"],"access":["desktop"],"deployment":["hosted"],"pricing":["subscription"],"targetUser":["creator"]},"citations":[{"url":"https://runway.com/changelog#2026-09-23","authority":"owner-controlled","publisher":"Runway"}],"score":1,"confidence":"high","correctionState":"current","publishedAt":"2026-09-23T00:00:00.000Z","effectiveAt":"2026-09-24T19:44:41.711Z","url":"https://wowsuchai.withglint.app/records/wsa_zf6wait4hxfw","apiUrl":"https://wowsuchai.withglint.app/api/v1/records/wsa_zf6wait4hxfw","webUrl":"https://wowsuchai.withglint.app/records/wsa_zf6wait4hxfw"},{"recordId":"wsa_i8vuawwn9f59","versionId":"ver_uICzZUgAm9Wf","eventType":"new_development","category":"application","title":"Astra-Ares","maker":"miuuyy","summary":"Astra-Ares is an experimental CLI tool that adapts reasoning effort for selected GPT-6 models during Codex tasks, using Jev to choose effort and how many generations it lasts. It is distributed as source and runs a separate, patched version of Codex CLI.","releaseNote":null,"classifications":{"developmentType":"application","capability":["agents","automation"],"modality":["text","code"],"access":["CLI"],"deployment":["hosted"],"pricing":["unknown"],"targetUser":["developer"]},"citations":[{"url":"https://github.com/miuuyy/Astra-Ares","authority":"owner-controlled","publisher":"GitHub new repositories"}],"score":1,"confidence":"high","correctionState":"current","publishedAt":"2026-09-22T14:10:07.000Z","effectiveAt":"2026-09-24T13:41:51.455Z","url":"https://wowsuchai.withglint.app/records/wsa_i8vuawwn9f59","apiUrl":"https://wowsuchai.withglint.app/api/v1/records/wsa_i8vuawwn9f59","webUrl":"https://wowsuchai.withglint.app/records/wsa_i8vuawwn9f59"},{"recordId":"wsa_9hgqk3kqjkxn","versionId":"ver_VBikGL1Jaoyu","eventType":"new_development","category":"application","title":"ZCode","maker":"Z.ai","summary":"ZCode is an AI coding workspace with desktop, browser, and terminal-agent interfaces. Its public repository contains the client, backend, shared UI, and Agent CLI and runtime source.","releaseNote":null,"classifications":{"developmentType":"application","capability":["code generation","agents"],"modality":["text","code"],"access":["web","CLI","desktop"],"deployment":["on-device"],"pricing":["unknown"],"targetUser":["developer"]},"citations":[{"url":"https://github.com/zai-org/ZCode","authority":"owner-controlled","publisher":"GitHub new repositories"}],"score":1,"confidence":"high","correctionState":"current","publishedAt":"2026-09-20T12:01:16.000Z","effectiveAt":"2026-09-24T13:41:20.492Z","url":"https://wowsuchai.withglint.app/records/wsa_9hgqk3kqjkxn","apiUrl":"https://wowsuchai.withglint.app/api/v1/records/wsa_9hgqk3kqjkxn","webUrl":"https://wowsuchai.withglint.app/records/wsa_9hgqk3kqjkxn"}],"corrections":[],"notes":[],"writer":{"kind":"model","model":"openai/gpt-6-luna"},"publishedAt":"2026-09-28T00:40:10.822Z","updatedAt":"2026-09-28T00:40:10.822Z","url":"https://wowsuchai.withglint.app/briefs/weekly/2026-W39","apiUrl":"https://wowsuchai.withglint.app/api/v1/briefs/weekly/2026-W39"},"nextCursor":null,"license":{"catalog":{"name":"CC BY 4.0","spdx":"CC-BY-4.0","url":"https://creativecommons.org/licenses/by/4.0/","attribution":"wowsuchai","attributionUrl":"https://wowsuchai.withglint.app","covers":["catalog metadata","classifications","generated summaries","feed event metadata"]},"excluded":{"covers":["source quotations","linked source content"],"terms":"Not licensed by wowsuchai. Source quotations and linked source content keep their original rights."}}}