AI tool releases — what shipped in open-source AI and ML

What this page is

A dated list of the versions actually published by 16 widely used open-source AI and ML repositories. The page was generated from the GitHub API on 2026-10-09 (UTC): the version tags, dates, licenses, star counts and links below are fields returned by the API, and each summary line is a quote from the release text. 59 release(s) fall inside the 45-day window. Nothing on this page was written by a language model, so there is no figure here that a model could have made up.

Release cycles differ by an order of magnitude: some of these projects cut several versions a month, others publish twice a year. That is the point of a dated list — it shows which tools are still moving and which ones you are depending on at a frozen version. 9 of the releases below are marked pre-release by their own maintainers.

Releases published in the last 45 days (59)

Version published	Repository	License	Stars	Release
plugin-ep-cuda/v0.3.0 — 2026-10-09	microsoft/onnxruntime	MIT	22,038	release
langchain==1.4.4 — 2026-10-08	langchain-ai/langchain	MIT	147,401	release
plugin-ep-cuda/v0.2.0 — 2026-10-08	microsoft/onnxruntime	MIT	22,038	release
b11514 — 2026-10-08	ggml-org/llama.cpp	MIT	130,554	release
langchain-core==1.6.9 — 2026-10-08	langchain-ai/langchain	MIT	147,401	release
b11513 — 2026-10-08	ggml-org/llama.cpp	MIT	130,554	release
mineru-4.0.11-released — 2026-10-08	opendatalab/MinerU	NOASSERTION	81,308	release
b11512 — 2026-10-08	ggml-org/llama.cpp	MIT	130,554	release
b11511 — 2026-10-08	ggml-org/llama.cpp	MIT	130,554	release
v0.40.2 — 2026-10-08	ollama/ollama	MIT	182,424	release
b11510 — 2026-10-08	ggml-org/llama.cpp	MIT	130,554	release
langchain-openai==1.7.0 — 2026-10-08	langchain-ai/langchain	MIT	147,401	release
v0.1.905-beta — 2026-10-08	unslothai/unsloth	Apache-2.0	77,534	release
langchain-core==1.6.8 — 2026-10-08	langchain-ai/langchain	MIT	147,401	release
v0.40.1 — 2026-10-07	ollama/ollama	MIT	182,424	release
langchain-huggingface==1.2.3 — 2026-10-07	langchain-ai/langchain	MIT	147,401	release
v0.1.904-beta — 2026-10-07	unslothai/unsloth	Apache-2.0	77,534	release
v1.9.5 — 2026-10-06	ggml-org/whisper.cpp	MIT	54,234	release
v0.1.903-beta — 2026-10-06	unslothai/unsloth	Apache-2.0	77,534	release
b5454 — 2026-10-06	ggml-org/whisper.cpp	MIT	54,234	release
v5.19.0 — 2026-10-06	huggingface/transformers	Apache-2.0	166,861	release
v0.41.0 — 2026-10-06	huggingface/diffusers	Apache-2.0	34,695	release
v0.39.0 — 2026-10-05	Comfy-Org/ComfyUI	GPL-3.0	136,467	release
v0.31.0 — 2026-10-05	vllm-project/vllm	Apache-2.0	93,416	release
v1.28.3 — 2026-10-02	microsoft/onnxruntime	MIT	22,038	release
v0.5.21 — 2026-10-02	sgl-project/sglang	Apache-2.0	36,893	release
v0.1.902-beta — 2026-10-01	unslothai/unsloth	Apache-2.0	77,534	release
v2.14.1 — 2026-09-30	pytorch/pytorch	NOASSERTION	103,912	release
v5.18.0 — 2026-09-30	huggingface/transformers	Apache-2.0	166,861	release
v0.38.0 — 2026-09-29	Comfy-Org/ComfyUI	GPL-3.0	136,467	release
v0.35.1 — 2026-09-29	ollama/ollama	MIT	182,424	release
mineru-4.0.10-released — 2026-09-29	opendatalab/MinerU	NOASSERTION	81,308	release
v0.32.3 — 2026-09-29	ml-explore/mlx	MIT	28,697	release
v0.35.0 — 2026-09-28	ollama/ollama	MIT	182,424	release
mineru-4.0.9-released — 2026-09-28	opendatalab/MinerU	NOASSERTION	81,308	release
v0.1.900-beta — 2026-09-28	unslothai/unsloth	Apache-2.0	77,534	release
mineru-4.0.8-released — 2026-09-27	opendatalab/MinerU	NOASSERTION	81,308	release
v0.40.0 — 2026-09-25	ollama/ollama	MIT	182,424	release
mineru-4.0.7-released — 2026-09-23	opendatalab/MinerU	NOASSERTION	81,308	release
v0.30.0 — 2026-09-22	vllm-project/vllm	Apache-2.0	93,416	release
v0.14.25 — 2026-09-21	run-llama/llama_index	MIT	52,445	release
v0.37.0 — 2026-09-21	Comfy-Org/ComfyUI	GPL-3.0	136,467	release
v0.5.20 — 2026-09-18	sgl-project/sglang	Apache-2.0	36,893	release
v0.36.0 — 2026-09-15	Comfy-Org/ComfyUI	GPL-3.0	136,467	release
1.19.0 — 2026-09-14	SillyTavern/SillyTavern	AGPL-3.0	34,233	release
v1.9.4 — 2026-09-11	ggml-org/whisper.cpp	MIT	54,234	release
b5130 — 2026-09-11	ggml-org/whisper.cpp	MIT	54,234	release
v1.30.0 — 2026-09-10	microsoft/onnxruntime	MIT	22,038	release
b5127 — 2026-09-10	ggml-org/whisper.cpp	MIT	54,234	release
v1.29.1 — 2026-09-10	microsoft/onnxruntime	MIT	22,038	release
v0.35.0 — 2026-09-09	Comfy-Org/ComfyUI	GPL-3.0	136,467	release
v5.17.0 — 2026-09-09	huggingface/transformers	Apache-2.0	166,861	release
v0.29.0 — 2026-09-09	vllm-project/vllm	Apache-2.0	93,416	release
v0.5.19 — 2026-09-05	sgl-project/sglang	Apache-2.0	36,893	release
v2.14.0 — 2026-09-02	pytorch/pytorch	NOASSERTION	103,912	release
v5.16.1 — 2026-08-26	huggingface/transformers	Apache-2.0	166,861	release
v5.16.0 — 2026-08-26	huggingface/transformers	Apache-2.0	166,861	release
v0.28.0 — 2026-08-26	vllm-project/vllm	Apache-2.0	93,416	release
v0.32.2 — 2026-08-25	ml-explore/mlx	MIT	28,697	release

What changed, in the projects' own words

Each line below is quoted from the beginning of the release body as returned by the API on 2026-10-09; markdown, code blocks, tables and links were stripped, contributor handles were removed, separate changelog entries are joined with a semicolon, and the text was cut at 200 characters. The wording is the maintainers', not ours.

microsoft/onnxruntime — plugin-ep-cuda/v0.3.0 — 2026-10-09

"This release brings sparse attention and packed continuous-batching support, INT2 and mixed-width QMoE execution, decode-kernel tuning, and reliability fixes to the ONNX Runtime CUDA Plugin EP."

microsoft/onnxruntime release plugin-ep-cuda/v0.3.0 — published 2026-10-09 — repository owner: microsoft — license MIT — 22,038 stars.

langchain-ai/langchain — langchain==1.4.4 — 2026-10-08

"Changes since langchain==1.4.3 release(langchain): 1.4.4 (#41163) fix(langchain): (SummarizationMiddleware) retry summary step on context overflow (#41159) chore(deps): bump fsspec from 2026.6.0 …"

langchain-ai/langchain release langchain==1.4.4 — published 2026-10-08 — repository owner: langchain-ai — license MIT — 147,401 stars.

microsoft/onnxruntime — plugin-ep-cuda/v0.2.0 — 2026-10-08

"This release brings expanded quantized inference and attention support, kernel performance improvements, and reliability fixes to the ONNX Runtime CUDA Plugin EP. Highlights include 2-bit …"

microsoft/onnxruntime release plugin-ep-cuda/v0.2.0 — published 2026-10-08 — repository owner: microsoft — license MIT — 22,038 stars.

ggml-org/llama.cpp — b11514 — 2026-10-08

"macOS Apple Silicon (arm64, KleidiAI enabled) DISABLED …"

ggml-org/llama.cpp release b11514 — published 2026-10-08 — repository owner: ggml-org — license MIT — 130,554 stars — flagged pre-release by the maintainers.

langchain-ai/langchain — langchain-core==1.6.9 — 2026-10-08

"Changes since langchain-core==1.6.8 feat(core): accept a callable in with_retry (#41158) …"

langchain-ai/langchain release langchain-core==1.6.9 — published 2026-10-08 — repository owner: langchain-ai — license MIT — 147,401 stars.

ggml-org/llama.cpp — b11513 — 2026-10-08

"CUDA: improve top-k algorithm selection (#28713); CUDA: radix top-k for large row counts; Replaces CUB's per-row DeviceTopKKernel with a grid-over-rows radix select, gated …"

ggml-org/llama.cpp release b11513 — published 2026-10-08 — repository owner: ggml-org — license MIT — 130,554 stars — flagged pre-release by the maintainers.

opendatalab/MinerU — mineru-4.0.11-released — 2026-10-08

"API：/v1/models 统一返回随服务版本变化的虚拟模型 ID（如 mineru-4.0.11），补充模型描述并保持注册时间稳定；/v1/tiers 的 current_model 同步使用该 ID，解析质量通过 tier 选择 / API: expose a single versioned virtual model ID through /v1/models, add its …"

opendatalab/MinerU release mineru-4.0.11-released — published 2026-10-08 — repository owner: opendatalab — license NOASSERTION — 81,308 stars.

ggml-org/llama.cpp — b11512 — 2026-10-08

"model : fix DFlash output head sharing (#30111); llama : fix DFlash output head sharing; dflash : read tied output weights from GGUF metadata; llama : share tied word embedding metadata; llama …"

ggml-org/llama.cpp release b11512 — published 2026-10-08 — repository owner: ggml-org — license MIT — 130,554 stars — flagged pre-release by the maintainers.

ggml-org/llama.cpp — b11511 — 2026-10-08

"CUDA: fix MMQ out-of-bounds reads (#29953) …"

ggml-org/llama.cpp release b11511 — published 2026-10-08 — repository owner: ggml-org — license MIT — 130,554 stars — flagged pre-release by the maintainers.

ollama/ollama — v0.40.2 — 2026-10-08

"README: add oxi to community integrations; server: hide duplicate and downgrade guards from list; a contributor made their first contribution …"

ollama/ollama release v0.40.2 — published 2026-10-08 — repository owner: ollama — license MIT — 182,424 stars — flagged pre-release by the maintainers.

ggml-org/llama.cpp — b11510 — 2026-10-08

"CUDA : looped PAD kernel for more than 65535 rows or slices (#30147) …"

ggml-org/llama.cpp release b11510 — published 2026-10-08 — repository owner: ggml-org — license MIT — 130,554 stars — flagged pre-release by the maintainers.

langchain-ai/langchain — langchain-openai==1.7.0 — 2026-10-08

"Changes since langchain-openai==1.6.7 release(openai): 1.7.0 (#41154) feat(openai): support decisions api (#41125) chore(deps): bump langgraph-sdk from 0.4.4 to 0.4.6 in /libs/partners/openai (#41096) …"

langchain-ai/langchain release langchain-openai==1.7.0 — published 2026-10-08 — repository owner: langchain-ai — license MIT — 147,401 stars.

unslothai/unsloth — v0.1.905-beta — 2026-10-08

"Turn any text or vision LLM into a Jev-style decision model in Unsloth, with decision accuracy going from 30% to 80%. Train, test, export and serve decision models directly from Unsloth. Also …"

unslothai/unsloth release v0.1.905-beta — published 2026-10-08 — repository owner: unslothai — license Apache-2.0 — 77,534 stars.

langchain-ai/langchain — langchain-core==1.6.8 — 2026-10-08

"Changes since langchain-core==1.6.7 fix(core): harden scoped and compatible IPv6 SSRF checks (#41151) …"

langchain-ai/langchain release langchain-core==1.6.8 — published 2026-10-08 — repository owner: langchain-ai — license MIT — 147,401 stars.

ollama/ollama — v0.40.1 — 2026-10-07

"server: proxy cloud usage and balance APIs; llama: fix clef head reads past 2GiB on windows; cmd: remove account step from CLI onboarding; manifest: avoid symlinks on Windows; docs: fix 6 dead links …"

ollama/ollama release v0.40.1 — published 2026-10-07 — repository owner: ollama — license MIT — 182,424 stars.

langchain-ai/langchain — langchain-huggingface==1.2.3 — 2026-10-07

"Changes since langchain-huggingface==1.2.2 fix(huggingface): use a supported Scaleway model in streaming test (#41095) chore(deps): bump langgraph-sdk from 0.4.4 to 0.4.6 …"

langchain-ai/langchain release langchain-huggingface==1.2.3 — published 2026-10-07 — repository owner: langchain-ai — license MIT — 147,401 stars.

unslothai/unsloth — v0.1.904-beta — 2026-10-07

"Turn any text or vision LLM into a Jev-style decision model in Unsloth, with decision accuracy going from 30% to 80%. Train, test, export and serve decision models directly from Unsloth. Also …"

unslothai/unsloth release v0.1.904-beta — published 2026-10-07 — repository owner: unslothai — license Apache-2.0 — 77,534 stars.

ggml-org/whisper.cpp — v1.9.5 — 2026-10-06

"Maintenance release. Updates to ggml v0.26.0. More info: dist : releases and versioning of ggml-org projects d1be6fde whisper : bump version to 1.9.5 (#4102) 4afec37b talk-llama : update llama.cpp …"

ggml-org/whisper.cpp release v1.9.5 — published 2026-10-06 — repository owner: ggml-org — license MIT — 54,234 stars — requested as ggerganov/whisper.cpp, the API now redirects it to ggml-org/whisper.cpp.

unslothai/unsloth — v0.1.903-beta — 2026-10-06

"This release adds a browser inside Unsloth (browser use coming very soon), so files, web pages and pages the model writes open right beside your chat. It also brings EmbeddingGemma 2, Google's new …"

unslothai/unsloth release v0.1.903-beta — published 2026-10-06 — repository owner: unslothai — license Apache-2.0 — 77,534 stars.

ggml-org/whisper.cpp — b5454 — 2026-10-06

This release was published with an empty description in the GitHub API; there is no changelog text to quote.

ggml-org/whisper.cpp release b5454 — published 2026-10-06 — repository owner: ggml-org — license MIT — 54,234 stars — requested as ggerganov/whisper.cpp, the API now redirects it to ggml-org/whisper.cpp — flagged pre-release by the maintainers.

huggingface/transformers — v5.19.0 — 2026-10-06

"EmbeddingGemma 2 is a multimodal embedding model from Google built on the Gemma 4 architecture. It encodes text, images, audio, and video, individually or combined in one input, into a shared …"

huggingface/transformers release v5.19.0 — published 2026-10-06 — repository owner: huggingface — license Apache-2.0 — 166,861 stars.

huggingface/diffusers — v0.41.0 — 2026-10-06

"This release brings Qwen-Image 2.1 to Diffusers, with text-to-image generation, image editing, native transparency, and LoRA training. Starting with this release, we’re adopting the Transformers …"

huggingface/diffusers release v0.41.0 — published 2026-10-06 — repository owner: huggingface — license Apache-2.0 — 34,695 stars.

Comfy-Org/ComfyUI — v0.39.0 — 2026-10-05

"Support LynnReal light Minimax-H3 vae; Update embedded docs to v0.5.13; Use higher quality defaults for Save Video encoding. feat: add DynamicGroup widget input; fix(assets): don't let a model …"

Comfy-Org/ComfyUI release v0.39.0 — published 2026-10-05 — repository owner: Comfy-Org — license GPL-3.0 — 136,467 stars — requested as comfyanonymous/ComfyUI, the API now redirects it to Comfy-Org/ComfyUI.

vllm-project/vllm — v0.31.0 — 2026-10-05

"This release features 717 commits from 307 contributors (96 new)! DeepSeek-V4.1-Flash performance: FlashMLA mega attention with the V4.1 NVFP4 compressed KV cache is now the SM100 default (#56935) …"

vllm-project/vllm release v0.31.0 — published 2026-10-05 — repository owner: vllm-project — license Apache-2.0 — 93,416 stars.

microsoft/onnxruntime — v1.28.3 — 2026-10-02

"This is a patch release on top of v1.28.2, containing memory-safety and model-validation fixes, a plugin execution fix, and packaging infrastructure updates. Fixed an out-of-bounds read …"

microsoft/onnxruntime release v1.28.3 — published 2026-10-02 — repository owner: microsoft — license MIT — 22,038 stars.

sgl-project/sglang — v0.5.21 — 2026-10-02

"779 PRs from 227 contributors. PD instances can switch between prefill and decode on the fly, no restart needed (#28403). The prefix cache now runs on a Rust core by default (#39627). DeepSeek-V4.1 …"

sgl-project/sglang release v0.5.21 — published 2026-10-02 — repository owner: sgl-project — license Apache-2.0 — 36,893 stars.

unslothai/unsloth — v0.1.902-beta — 2026-10-01

"This release brings faster navigation, shareable run settings, and clearer errors to Unsloth Desktop. It also 4x speeds up Laya decisions, expands hosted Decision API support, and keeps NVFP4, INT4 …"

unslothai/unsloth release v0.1.902-beta — published 2026-10-01 — repository owner: unslothai — license Apache-2.0 — 77,534 stars.

pytorch/pytorch — v2.14.1 — 2026-09-30

"This release is meant to fix the following regressions and silent correctness issues: Fix incorrect torch.linalg.lstsq solutions on MPS for complex batched underdetermined systems (#196113), fixed …"

pytorch/pytorch release v2.14.1 — published 2026-09-30 — repository owner: pytorch — license NOASSERTION — 103,912 stars.

huggingface/transformers — v5.18.0 — 2026-09-30

"Nemotron 3 Diarization is an open-weight streaming speaker diarization model designed to determine 'who spoke when' in real-world audio. It supports both streaming and offline inference, handles up …"

huggingface/transformers release v5.18.0 — published 2026-09-30 — repository owner: huggingface — license Apache-2.0 — 166,861 stars.

Comfy-Org/ComfyUI — v0.38.0 — 2026-09-29

"Improve qwen 2.1 KV cache location logic; Compile Qwen Image 2.1 transformer blocks; Allow model files to contain which attention should be used for each block. JsonExtractString can now handle JSON …"

Comfy-Org/ComfyUI release v0.38.0 — published 2026-09-29 — repository owner: Comfy-Org — license GPL-3.0 — 136,467 stars — requested as comfyanonymous/ComfyUI, the API now redirects it to Comfy-Org/ComfyUI.

ollama/ollama — v0.35.1 — 2026-09-29

"Ollama now supports Clef and Clef Flash, Cloudflare's new open-source decision models, through /v1/systemone. Clef (27B) and Clef Flash (9B) are multimodal: requests can now include images alongside …"

ollama/ollama release v0.35.1 — published 2026-09-29 — repository owner: ollama — license MIT — 182,424 stars.

opendatalab/MinerU — mineru-4.0.10-released — 2026-09-29

"WebUI：非安全上下文（如 HTTP 或非 localhost 访问）下 crypto.randomUUID 不可用时提供 UUID 回退实现，避免预览等功能失效 / WebUI: fall back to an alternative UUID implementation when crypto.randomUUID is unavailable in insecure contexts …"

opendatalab/MinerU release mineru-4.0.10-released — published 2026-09-29 — repository owner: opendatalab — license NOASSERTION — 81,308 stars.

ml-explore/mlx — v0.32.3 — 2026-09-29

"Fix scan and sort for a zero-size axis; Add a correction parameter in std and var; Fix tests/run.py stuck in macOS CI; Fix sorted gather_qmm NAX row overflow above 32K; Fix deadlock caused …"

ml-explore/mlx release v0.32.3 — published 2026-09-29 — repository owner: ml-explore — license MIT — 28,697 stars.

ollama/ollama — v0.35.0 — 2026-09-28

"Ollama now supports decision models through /v1/systemone, based on TypeSafe’s Jev API. Decision models return choices, probabilities, and scores instead of text. Use them for tasks such as ticket …"

ollama/ollama release v0.35.0 — published 2026-09-28 — repository owner: ollama — license MIT — 182,424 stars.

opendatalab/MinerU — mineru-4.0.9-released — 2026-09-28

"PDF 原生文本提取：重建词间空格，修复空格丢失导致的英文单词黏连，并保护代码 span 与相邻文本的间距 / Native PDF text extraction: rebuild inter-word spacing to fix lost spaces merging English words, with code-span spacing protection; PDF render …"

opendatalab/MinerU release mineru-4.0.9-released — published 2026-09-28 — repository owner: opendatalab — license NOASSERTION — 81,308 stars.

unslothai/unsloth — v0.1.900-beta — 2026-09-28

"We're adding support for decision models, a unified Library for docs and media, document viewer, many Apple Silicon improvements, creation of Skills, and ~4.5× faster image and video generation. Run …"

unslothai/unsloth release v0.1.900-beta — published 2026-09-28 — repository owner: unslothai — license Apache-2.0 — 77,534 stars.

opendatalab/MinerU — mineru-4.0.8-released — 2026-09-27

"PDF 原生内核：跨分析窗口复用文档渲染会话与文本快照，原生 span 内容与赋值直接消费、免去字符级重组 / Native PDF kernel: reuse document render sessions and owned text snapshots across analysis windows; consume native span content and assignments …"

opendatalab/MinerU release mineru-4.0.8-released — published 2026-09-27 — repository owner: opendatalab — license NOASSERTION — 81,308 stars.

ollama/ollama — v0.40.0 — 2026-09-25

"Models run on MLX on Apple Silicon by default In this release, on Apple Silicon devices, model architectures supported by the MLX runtime will automatically run on MLX. Additional models include …"

ollama/ollama release v0.40.0 — published 2026-09-25 — repository owner: ollama — license MIT — 182,424 stars.

opendatalab/MinerU — mineru-4.0.7-released — 2026-09-23

"MHTML/MHT 文件支持：解析与源文件在线预览 / MHTML/MHT file support: parsing and in-app source preview; MHTML 预览稳定性：保留归档样式表状态与 base URL、本地 CSS fragment URL、预加载重定向恢复、预览内链接可用 / MHTML preview stability: preserve …"

opendatalab/MinerU release mineru-4.0.7-released — published 2026-09-23 — repository owner: opendatalab — license NOASSERTION — 81,308 stars.

vllm-project/vllm — v0.30.0 — 2026-09-22

"This release features 762 commits from 315 contributors (104 new)! New models: DeepSeek-V4.1-Flash (#56214, #56228, #56208) with the whole KV stored in MXFP8 through the FlashMLA V4.1 record on SM100 …"

vllm-project/vllm release v0.30.0 — published 2026-09-22 — repository owner: vllm-project — license Apache-2.0 — 93,416 stars.

run-llama/llama_index — v0.14.25 — 2026-09-21

"llama-index-agent-agentmesh [0.3.0]; fix: resolve a ton of security alerts (#22855); llama-index-agent-azure [0.4.0]; fix: resolve a ton of security alerts (#22855); llama-index-callbacks-argilla …"

run-llama/llama_index release v0.14.25 — published 2026-09-21 — repository owner: run-llama — license MIT — 52,445 stars.

Comfy-Org/ComfyUI — v0.37.0 — 2026-09-21

"Aimdo 0.5.5 + Auto-detect and enable --fast-disk when the disk is fast (CORE-440); [Partner Nodes] feat(OpenAI): add transparent background support for GPT Image 2; Add CFG control to YuE2 Generate …"

Comfy-Org/ComfyUI release v0.37.0 — published 2026-09-21 — repository owner: Comfy-Org — license GPL-3.0 — 136,467 stars — requested as comfyanonymous/ComfyUI, the API now redirects it to Comfy-Org/ComfyUI.

sgl-project/sglang — v0.5.20 — 2026-09-18

"713 PRs from 237 contributors. New models in this release (see the cookbook for all supported models): Sampling masks for RL rollouts. With return_sampling_mask, each decode step returns the exact …"

sgl-project/sglang release v0.5.20 — published 2026-09-18 — repository owner: sgl-project — license Apache-2.0 — 36,893 stars.

Comfy-Org/ComfyUI — v0.36.0 — 2026-09-15

"Add new model blueprints and reorganize subgraph categories; main: bump the AMD Windows VA quota to 4TB (CORE-409); [Partner Noes] feat(OpenRouter): add Microsoft mai-image-2.6 models; quant_ops …"

Comfy-Org/ComfyUI release v0.36.0 — published 2026-09-15 — repository owner: Comfy-Org — license GPL-3.0 — 136,467 stars — requested as comfyanonymous/ComfyUI, the API now redirects it to Comfy-Org/ComfyUI.

SillyTavern/SillyTavern — 1.19.0 — 2026-09-14

"Anthropic: Added support for Claude Fable 5 and 5.1, Claude Opus 4.8 and 5, Claude Sonnet 5. OpenAI: Added support for GPT-5.6 models, GPT-6 Astra. Other providers: Added support for Gemini …"

SillyTavern/SillyTavern release 1.19.0 — published 2026-09-14 — repository owner: SillyTavern — license AGPL-3.0 — 34,233 stars.

ggml-org/whisper.cpp — v1.9.4 — 2026-09-11

"New version has been released. More info: dist : releases and versioning of ggml-org projects 927cfce3 metal : remove leftover ggml-metal.metal kernels file (#4051) dd80eb2a scripts : use sed instead …"

ggml-org/whisper.cpp release v1.9.4 — published 2026-09-11 — repository owner: ggml-org — license MIT — 54,234 stars — requested as ggerganov/whisper.cpp, the API now redirects it to ggml-org/whisper.cpp.

ggml-org/whisper.cpp — b5130 — 2026-09-11

This release was published with an empty description in the GitHub API; there is no changelog text to quote.

ggml-org/whisper.cpp release b5130 — published 2026-09-11 — repository owner: ggml-org — license MIT — 54,234 stars — requested as ggerganov/whisper.cpp, the API now redirects it to ggml-org/whisper.cpp — flagged pre-release by the maintainers.

microsoft/onnxruntime — v1.30.0 — 2026-09-10

"ONNX Runtime 1.30.0 expands generative AI inference, improves CPU and GPU performance, adds Go bindings, and strengthens runtime reliability. These notes cover changes since ONNX Runtime 1.29.1."

microsoft/onnxruntime release v1.30.0 — published 2026-09-10 — repository owner: microsoft — license MIT — 22,038 stars.

ggml-org/whisper.cpp — b5127 — 2026-09-10

"Note: the Windows arm64 (CUDA) build (whisper-bin-win-cuda-*-arm64.zip) uses a preview edition of the CUDA Toolkit for Windows on Arm and should be considered experimental."

ggml-org/whisper.cpp release b5127 — published 2026-09-10 — repository owner: ggml-org — license MIT — 54,234 stars — requested as ggerganov/whisper.cpp, the API now redirects it to ggml-org/whisper.cpp — flagged pre-release by the maintainers.

microsoft/onnxruntime — v1.29.1 — 2026-09-10

"This is a patch release on top of v1.29.0, containing GroupQueryAttention capability and KV-cache layout improvements, plugin Execution Provider performance tooling updates, and targeted graph …"

microsoft/onnxruntime release v1.29.1 — published 2026-09-10 — repository owner: microsoft — license MIT — 22,038 stars.

Comfy-Org/ComfyUI — v0.35.0 — 2026-09-09

"[Partner Nodes] chore(Google): drop retiring Veo 2 and Veo 3.0 models; [Partner Nodes] feat(Recraft): add V4 Styles; [Partner Nodes] feat(WAN): add WAN3-Prime model support; Bump …"

Comfy-Org/ComfyUI release v0.35.0 — published 2026-09-09 — repository owner: Comfy-Org — license GPL-3.0 — 136,467 stars — requested as comfyanonymous/ComfyUI, the API now redirects it to Comfy-Org/ComfyUI.

huggingface/transformers — v5.17.0 — 2026-09-09

"Hy4-Preview is a 780B-parameter mixture-of-experts language model that activates 49B parameters per token. Each MoE layer holds 256 routed experts plus one always-active shared expert and routes every …"

huggingface/transformers release v5.17.0 — published 2026-09-09 — repository owner: huggingface — license Apache-2.0 — 166,861 stars.

vllm-project/vllm — v0.29.0 — 2026-09-09

"This release features 594 commits from 277 contributors (91 new)! Model Runner V2 is now the default for all models (#53183), completing the rollout that began with pooling models (#48290). MRV2 also …"

vllm-project/vllm release v0.29.0 — published 2026-09-09 — repository owner: vllm-project — license Apache-2.0 — 93,416 stars.

sgl-project/sglang — v0.5.19 — 2026-09-05

"786 PRs from 214 contributors. New models in this release (see the cookbook for all supported models): GLM-5.3 deployment guide: link; PaddleOCR-VL deployment guide: link; Kimi-K2.7-Code-MXFP4 …"

sgl-project/sglang release v0.5.19 — published 2026-09-05 — repository owner: sgl-project — license Apache-2.0 — 36,893 stars.

pytorch/pytorch — v2.14.0 — 2026-09-02

"NVGEMM brings CuTeDSL-generated CUTLASS kernels to Inductor, with epilogue fusion, scaled and NVFP4 GEMM, and grouped-reduction epilogues autotuned alongside Triton and ATen torch.switch generalizes …"

pytorch/pytorch release v2.14.0 — published 2026-09-02 — repository owner: pytorch — license NOASSERTION — 103,912 stars.

huggingface/transformers — v5.16.1 — 2026-08-26

"This is a special release as we include GLM! (and a few small fixes) GLM-5.3-Flash, the first natively multimodal model in the GLM-5 series. With 320B total parameters and just 18B active parameters …"

huggingface/transformers release v5.16.1 — published 2026-08-26 — repository owner: huggingface — license Apache-2.0 — 166,861 stars.

huggingface/transformers — v5.16.0 — 2026-08-26

"Qwen4-Exp builds on Qwen3.5's hybrid text and multimodal architecture with three key components: GatedResidual (GR), Qwen Sparse Attention (QSA), and Per-Layer Embedding (PLE). GR is a Qwen-developed …"

huggingface/transformers release v5.16.0 — published 2026-08-26 — repository owner: huggingface — license Apache-2.0 — 166,861 stars.

vllm-project/vllm — v0.28.0 — 2026-08-26

"This release features 584 commits from 270 contributors (76 new)! Kimi-K3 performance push: a major optimization effort for Kimi-K3 across the stack — Decode Context Parallel (DCP) support (#50484) …"

vllm-project/vllm release v0.28.0 — published 2026-08-26 — repository owner: vllm-project — license Apache-2.0 — 93,416 stars.

ml-explore/mlx — v0.32.2 — 2026-08-25

"Preserve subnormal float values when casting to bool; Support assigning through a bare Ellipsis index; Fix divmod truncating the quotient for floats; Add force_fused option …"

ml-explore/mlx release v0.32.2 — published 2026-08-25 — repository owner: ml-explore — license MIT — 28,697 stars.

How this page is built

• Source: the GitHub REST API, endpoints /repos/<owner>/<repo> and /repos/<owner>/<repo>/releases?per_page=5, read on 2026-10-09 (UTC). No other source is used, and no field is estimated.
• Scope: 16 repositories that are widely used in AI and ML and that the API confirmed as reachable. A release is kept only if its published_at field falls inside the last 45 days.
• Generated by pipelines/releases_page.py, which fetches, filters, sorts and quotes by script. No language model is called, so no summary on this page is a paraphrase: each one is the maintainers' own text, truncated.
• Guard: the script refuses to overwrite this page when fewer than 3 releases are collected, or fewer than 50% of the previous run's count, so a GitHub outage cannot erase the page or make it look as though every project went quiet.
• Every run is recorded in data/history/releases/<date>.json so today's list can be diffed against tomorrow's.
• License and star figures are the values returned by the API on 2026-10-09 and move over time; re-check them before you rely on them. Star counts are API-reported, not audited.
• Releases seen but not listed: 21, outside the 45-day window, undated drafts, or from a repository whose API call failed. They are kept in the dated history file rather than dropped.

Page assembled 2026-10-09 (UTC) by FTL Lab from the GitHub API. Written at /Volumes/Cascade/projects/active/BlogIA_project/content/en/releases/releases.md by pipelines/releases_page.py.