Ollama – v0.30.0-rc31

Ollama v0.30.0-rc31 🦙

If you aren’t running Ollama yet, you are missing out on the absolute easiest way to host massive LLMs like Llama 3 or DeepSeek-R1 right on your own hardware. It’s the ultimate playground for privacy-focused tinkering and building local dev workflows without worrying about API costs or data leaks!

This latest release candidate is a targeted stability update specifically designed to keep things running smoothly for the Apple Silicon crowd. Here’s the breakdown:

• MLX Checkout Fix: Includes a critical CI (Continuous Integration) fix for non-shallow MLX checkouts.
• Enhanced Stability: For those leveraging Apple’s MLX framework to squeeze every bit of performance out of their Mac, this ensures the build process is much more reliable and consistent.

It’s a small but mighty update to ensure the plumbing stays solid for high-performance local inference! 🛠️

🔗 View Release

Ollama – v0.30.0-rc30

Ollama v0.30.0-rc30 🦙

If you aren’t running Ollama yet, you are missing out on the ultimate playground for local LLMs! It is hands-down the easiest way to pull models like Llama 3, DeepSeek-R1, and Mistral directly onto your machine without worrying about subscription fees or data privacy. Whether you’re a dev building local integrations via their REST API or just a tinkerer playing with GGUF files, Ollama is the backbone of the local AI movement.

This latest update is v0.30.0-rc30, a Release Candidate build that is getting us closer to the next major stable milestone. Here is the breakdown:

• Stability Polish: This is a targeted maintenance release designed to iron out the kinks discovered in previous RC iterations.
• Bug Squashing: The focus here is on refining the local inference engine to ensure everything runs snappy and reliably before the full stable rollout.
• Pre-release Testing: Since this is an `rc` build, it’s the perfect time for us power users to test the waters and ensure our local workflows are rock solid.

Keep your eyes peeled—these RC builds are where the final magic happens! 🛠️

🔗 View Release

Ollama – v0.30.0-rc29

Ollama v0.30.0-rc29 is here! 🛠️

If you aren’t running Ollama yet, you’re missing out on the easiest way to host heavy-hitters like Llama 3, DeepSeek-R1, and Mistral directly on your own hardware. It’s the ultimate toolkit for local LLM privacy, giving you a full CLI and REST API to integrate models into your own dev projects without ever hitting an external server.

This latest release is a Release Candidate (rc29), which means the team is in the final stretch of polishing the upcoming v0.30.0 stable launch. Here’s what’s happening in this build:

• Stability Refinements: This version is laser-focused on squashing bugs discovered during the testing phase to ensure a rock-solid experience once the stable version drops.
• Performance Tweaks: Expect smoother model loading and faster inference speeds as they continue to optimize how the engine talks to your hardware backends.
• Community Testing: As an RC build, it’s the perfect playground for us tinkerers to stress-test upcoming architectural changes before they go mainstream.

If you want to stay on the bleeding edge of local AI performance, grab this candidate build and help shake out those last few bugs! 🚀

🔗 View Release

Ollama – v0.30.0-rc28

Ollama v0.30.0-rc28 🛠️

If you’re running local LLMs, you know Ollama is the go-to for getting models like Llama 3 and DeepSeek-R1 up and running effortlessly on your machine. This latest release candidate brings a specific tweak to how the engine handles integrated graphics!

What’s New:

• iGPU Management Update: The developers have largely disabled iGPUs (integrated graphics) by default in this build to prioritize stability.

New Environment Variable: A new flag, `OLLAMA_IGPU_ENABLE`, has been added. This is a lifesaver if you actually want* to utilize that integrated graphics power—you can now manually toggle it back on!

This is a great move if you want to ensure your heavy lifting stays strictly on your dedicated GPU, but definitely worth checking out if you’re working on a laptop where every bit of VRAM counts. 🚀

🔗 View Release

ComfyUI – v0.22.3

ComfyUI v0.22.3 is live! 🚀

If you haven’t dived into this node-based powerhouse yet, you are missing out on some serious granular control over your Stable Diffusion workflows. It’s the ultimate playground for anyone who loves connecting nodes to build complex, custom image and video generation pipelines. 🎨

What’s new in v0.22.3:

• Stability & Maintenance: This is a minor patch release focused on keeping your engine running reliably.
• Bug Fixes & Refinement: This update targets smoother node execution and squashing small UI glitches to ensure your creative flow stays buttery smooth. 🛠️

Time to check for updates and keep those custom nodes humming!

🔗 View Release

Ollama – v0.30.0-rc27

Ollama v0.30.0-rc27 🛠️

If you’re running local LLMs, you know Ollama is the go-to for getting models like Llama 3 and DeepSeek-R1 up and running with zero friction. This latest release candidate is a targeted update focused on stability and build fixes!

What’s new in this release:

• Windows Path Fix: The team implemented a crucial workaround for Windows paths during CPU builds. If you’ve ever tried compiling Ollama from source on Windows, this should help prevent those pesky build errors! 🪟
• CI Improvements: This update includes continuous integration (CI) tweaks to ensure smoother, more reliable automated testing cycles for the whole community.

It’s a small but mighty update aimed at making the development pipeline more robust for everyone contributing to the project! 🚀

🔗 View Release

Ollama – v0.30.0-rc26: Merge remote-tracking branch ‘upstream/main’ into llama-runner-phase-0

Ollama v0.30.0-rc26 is officially hitting the staging area! 🛠️

If you haven’t jumped on the Ollama train yet, this is the ultimate toolkit for running powerful LLMs like Llama 3, DeepSeek-R1, and Mistral locally on your own hardware with zero friction. It’s a game-changer for anyone wanting to experiment with AI without relying on expensive cloud APIs.

This latest release candidate is all about strengthening the internal engine. Here is what’s happening under the hood:

• Llama Runner Optimization: A major merge of the `upstream/main` branch into the `llama-runner-phase-0` development branch has taken place. This indicates some heavy lifting is happening to optimize how the runner handles model execution and performance.
• Server Stability Improvements: The team has resolved specific conflicts within `server/images.go` and its associated tests. For us users, this means a more stable experience when managing, pulling, and storing model images on your local machine.

While it’s still in the RC stage, these backend refinements are paving the way for a much smoother and more performant local AI workflow! 🚀

🔗 View Release

Ollama – v0.30.0-rc25

Ollama just dropped v0.30.0-rc25, and it looks like the team is hard at work polishing up this latest release candidate! 🛠️

If you aren’t using Ollama yet, it’s the ultimate toolkit for running powerful LLMs like Llama 3, DeepSeek-R1, and Mistral directly on your own hardware. It takes all the headache out of downloading and configuring models, giving you a local API and CLI to play with.

What’s new in this release:

• Windows on Arm (WoA) Improvements: Huge win for devs using ARM-based Windows laptops! This update includes specific fixes for WoA cross-compilation, making it much smoother to run local LLMs on specialized hardware. 💻
• CI Pipeline Fixes: The team has implemented critical fixes for the Continuous Integration pipeline to ensure much more stable builds moving forward.

This is a targeted stability update designed to iron out the kinks as we approach the next full stable release. Keep those local models running smooth! 🚀

🔗 View Release

Tater – Tater v84

🥔 Tater v84 — “Runs On Whatever Weird Thing You Own” 😏💻

If you’ve been looking for a way to run a powerful, local-native AI assistant without sacrificing privacy, Tater is your new best friend. Powered by the Hydra four-headed planning engine (Astraeus, Thanatos, Minos, and Hermes), it handles everything from perception to speech using your own local LLMs via Ollama or LM Studio. Whether you’re automating your Home Assistant setup or chatting via Discord or Telegram, Tater keeps your data exactly where it belongs: with you.

The v84 update is all about embracing “Frankenstein” hardware! This release focuses on making Tater run smoothly on everything from high-end rigs to that old Linux box gathering dust in the corner.

⚙️ New Setup Profiles

Say goodbye to manual configuration headaches. The new `setup_tater.sh` script now includes hardware-specific profiles for:

• CPU & macOS
• NVIDIA GPUs
• AMD (ROCm / Strix Halo)
• NVIDIA Jetson (Thor)

🧠 Smarter Hardware Detection

Tater is getting much sharper at detecting and routing acceleration paths. This ensures models like Kokoro, SpeechBrain, and Faster Whisper behave predictably across various backends, including Apple Metal/MPS, MLX, AMD ROCm, and NVIDIA CUDA.

🖥️ Cleaner Runtime Startup

The `run_ui.sh` script has been streamlined to reduce friction. It now auto-loads your runtime profile settings and automatically utilizes your `.venv`. Less “startup nonsense,” more time actually using your potato! 🥔

🔗 View Release

Ollama – v0.30.0-rc24

Ollama v0.30.0-rc24 🛠️

If you’re running local LLMs, you know Ollama is the absolute go-to for getting models like Llama 3 and DeepSeek-R1 up and running with zero friction. It’s basically your personal command center for hosting powerful open-source models right on your own hardware.

This latest release candidate (rc24) is all about the home stretch! We are seeing a minor version bump that signals the devs are polishing the engine and squashing bugs to prepare us for the official stable v0.30.0 launch.

What’s happening in this update:

• Stability Focus: This specific RC update is focused on fine-tuning performance and ensuring everything is rock-solid for the upcoming major version.
• Bug Fixes & Refinement: The team is working through crucial testing to ensure compatibility with the latest model weights before the full rollout.

Keep an eye on this one—major version bumps in Ollama often bring massive improvements to concurrency and memory management, which is huge for anyone trying to run multiple models or larger parameters! 🚀

🔗 View Release