Ollama v0.40.11 adds macOS speech generation and vLLM integration
ByPulseAugur Editorial·[12 sources]·
Ollama has released version v0.40.11, introducing several new features and improvements. The update enables speech generation on macOS, allowing users to play text or model replies aloud and stream audio via an OpenAI-compatible endpoint. Additionally, Ollama now integrates with vLLM Semantic Router's decision models, enabling classification, scoring, and yes/no answers for various inputs. The release also includes updates to specific models like qwen3coder and improvements to llama.cpp and mlxrunner.
AI
IMPACT
Enhances Ollama's utility with new audio capabilities and improved decision-making integrations.
RANK_REASON
Software release notes for a tool, not a frontier model or research paper.
<h2>Speech</h2> <p>Ollama can now generate <a href="https://docs.ollama.com/capabilities/speech" rel="nofollow">speech</a> on macOS. <code>ollama run</code> plays text or a chat model's reply aloud as the audio is generated, and apps can stream audio through the OpenAI-compatible…
Pretrained raster-scan autoregressive image models can be adapted into flexible any-order generators via future-aware positional embeddings and an ordering curriculum, enabling inpainting and outpainting with minimal extra compute.
arXiv cs.LG
TIER_1English(EN)·Shixuan Liu, Cheng Tang, Yuzheng Hu, Fan Wu, Han Zhao, Jiaqi W. Ma·
arXiv:2610.28883v1 Announce Type: new Abstract: Online reinforcement learning (RL) has achieved significant success in decision-making tasks, but modern online RL remains highly sensitive to the quality of training experience. Understanding the contribution of individual training…
Self-forcing training with teacher distillation and asymmetric conditioning stabilizes multi-turn image editing against accumulating distribution shifts.
arXiv:2601.01535v2 Announce Type: replace Abstract: Flexible image tokenizers aim to represent an image using an ordered 1D variable-length token sequence. This flexible tokenization is typically achieved through nested dropout, where a portion of trailing tokens is randomly trun…
arXiv:2610.34333v1 Announce Type: new Abstract: Language models switch tasks from demonstrations, but vision still often relies on separate systems for restoration, editing, and stylization. A before--after example offers a shared instruction only if the transferable change can b…
arXiv cs.CV
TIER_1English(EN)·Guangyu Meng, Pengfei Gu, Peixian Liang, John P. Lalor, Erin Wolf Chambers, Danny Z. Chen·
arXiv:2610.32884v1 Announce Type: new Abstract: Deep clustering offers a promising approach for analyzing large unlabeled medical image datasets, yet prevailing methods still face two limitations: they may neglect class-agnostic structural priors that capture object boundaries an…
<p>Cloudflare's <code>security-audit-skill</code> has 45,592 stars on GitHub. The Chinese translation was published today and has zero. That gap is the whole reason the translation exists — not because the original is hard to read for everyone, but because a security workflow is …
