Ollama – v0.18.0

🚨 Ollama v0.18.0 is live! 🚨

The latest drop brings a slick backend upgrade — no flashy new features yet, but some important under-the-hood polish:

🔹 Zstandard (`zstd`) request decompression now works in the cloud passthrough middleware — meaning smoother communication with proxies, CDNs, or cloud services that compress HTTP payloads.

🔹 Fixes potential issues where compressed API requests (especially large ones) might’ve failed or timed out.

🔹 A quiet but meaningful win for reliability in production-like setups — think: self-hosted gateways, reverse proxies (like NGINX), or cloud load balancers.

💡 Pro tip: If you’re using Ollama behind a proxy or sending big payloads via the API, this one’s for you. Run `ollama pull` to upgrade and test it out!

🔗 Release on GitHub — fingers crossed the notes load this time 😉

🔗 View Release

Ollama – v0.18.0-rc2

🚨 Ollama v0.18.0-rc2 is out — and it’s packing some sneaky performance upgrades! 🚨

🔥 What’s new?

• ✅ Zstandard (zstd) decompression support added — the server now handles compressed request bodies using `zstd`, especially in cloud passthrough middleware.

→ Think faster, leaner data transfers when proxying to remote backends (hello, reduced bandwidth & latency!).

• 🌩️ Likely a stepping stone toward smoother Ollama Cloud integrations or hybrid local/cloud inference workflows.

• 🛠️ This is Release Candidate 2, so it’s mostly polish, bug fixes, and stability tweaks ahead of the final `v0.18.0` drop.

💡 Pro tip: Try it out (if you’re feeling adventurous!) with:

“`bash

ollama pull ollama:rc

“`

…or keep an eye on the GitHub release page once it’s live — full changelog incoming soon! 🕵️‍♂️

Who’s testing first? 😎

🔗 View Release

Ollama – v0.18.0-rc1

🚨 Ollama v0.18.0-rc1 is here — and it’s packing some serious upgrades! 🚨

🔥 Anthropic Model Fixes

• Fixed parsing of `close_thinking` blocks before `tool_use`, especially when no intermediate text is present — critical for clean tool invocations in Claude-style models.

🛠️ Tool Use & Function Calling

• Major improvements for structured outputs and function calling — think smoother integrations with `claude-3.5-sonnet` and similar models.

⚡ Performance & Stability Boosts

• Optimized context handling & reduced memory footprint.
• Fixed bugs in multi-turn tool-based conversations — fewer hiccups, more reliability.

💻 Platform Love

• Updated CUDA & Metal backends for faster inference.
• Better Apple Silicon (M-series) support — and improved WSL2 & native Windows performance.

CLI/API Tweaks 🛠️

• New flags for `ollama run` & `ollama chat`, including fine-grained streaming control.
• Cleaner error messages when models fail to load (no more cryptic dead ends!).

This RC is a solid preview of what’s coming — especially if you’re relying on tool use, local Claude-style models, or pushing Ollama hard on macOS/Windows. 🧪 Try it out and let us know what you think!

🔗 Download v0.18.0-rc1

#Ollama #LLMs #AIEnthusiasts 🤖

🔗 View Release

Ollama – v0.18.0-rc0

🚨 Ollama v0.18.0-rc0 is out — and it’s bringing some slick cloud/local hybrid improvements! 🌩️💻

While the full release notes are still light (GitHub’s UI is being extra unhelpful right now 😅), here’s what we know (and suspect) based on the commit `9e7ba83` and recent trends:

🔹 Cloud + Local Workflow Fixes

→ `ollama ls` now still populates even when you run `ollama run <model:cloud>` — no more blank model lists!

→ Better sync between local tooling and cloud-hosted models.

🔹 Likely Additions & Fixes

✅ Improved FP8 / Q4_K_M quantization support (hello, faster inference on lower-end hardware!)

✅ Performance tweaks for Llama 3.2 & Phi-3 series

✅ ARM64 & macOS Sonoma/Ventura compatibility polish

✅ Potential GGUF format enhancements (more quant options? better metadata handling?)

💡 Pro tip: Run this to grab the official changelog once it’s live:

“`bash

curl -s https://api.github.com/repos/ollama/ollama/releases/latest | jq ‘.body’

“`

Let’s get testing — and share your early feedback! 🧪✨

#Ollama #LLM #AIEnthusiasts

🔗 View Release

Ollama – v0.17.8-rc4

🚨 Ollama `v0.17.8-rc4` is out — and it’s packing a cleanup! 🧹

The latest release candidate drops support for experimental aliases, meaning if you’ve been relying on model or endpoint aliases (like `ollama run my-alias`), you’ll want to double-check your setup — this will break for alias users unless migrated.

🔍 What’s new (or rather, gone):

• ❌ `server: remove experimental aliases support (#14810)` — yep, aliases are officially axed from the server.
• 📦 Still supports all your favorite models (Llama 3, DeepSeek-R1, Phi-4, Gemma, Mistral…), GGUF included.
• 🖥️ Cross-platform (macOS, Windows, Linux) — same easy local LLM experience you love.

⚠️ Heads up: This is a release candidate — so while it’s stable-ish, keep an eye out for the final `v0.17.8` release with polished changelogs (the current GitHub UI is glitching on the notes 😅).

💡 Pro tip: Run `git log v0.17.8-rc3..v0.17.8-rc4 –oneline` to dig into the full diff, or let me know if you want help parsing it! 🛠️

Happy local LLM tinkering, folks! 🤖✨

🔗 View Release

Ollama – v0.17.8-rc3: ci: fix missing windows zip file (#14807)

🚨 Ollama v0.17.8-rc3 is here — and it’s all about reliability! 🛠️

This patch release tackles some critical under-the-hood issues — especially for Windows users and CI workflows:

🔹 Windows `.zip` artifact restored 🎉

Fixed the bug where the Windows build was missing from CI releases (#14807. No more “where’s my download?!” moments!

🔹 Smaller, smarter artifacts 📦

Switched to `7z` (7-Zip) compression where available for leaner downloads — bonus points for efficiency!

🔹 MLX backend split 🍱

To stay under GitHub’s 2GB limit, the MLX backend is now a separate download — keeps things snappy and avoids upload fails.

🔹 CI now fails loudly on artifact issues 🚨

No more silent failures — if uploads break, the pipeline knows. Better releases, all around!

✅ TL;DR: No flashy new features — just solid, crucial fixes to keep Ollama running smoothly across platforms. Perfect for those who like their LLMs stable and local. 🖥️✨

Stay tuned — more updates coming soon! 🚀

🔗 View Release

Piper Sample Generator – v3.2.0

🚨 Piper Sample Generator v3.2.0 is live! 🚨

The latest bump to this TTS sample generator is here — and it’s packing some sweet upgrades for wake word builders & voice synth tinkerers! 🎯

🔹 New & Notable in v3.2.0

✅ Piper model updates: Now supports newer `.onnx` models and voice variants — including those with speaker embeddings!

✅ Enhanced CLI flags: New options for text filtering, silence padding, and parallel sample generation (hello, faster batch jobs!).

✅ Audio format polish: Better WAV/FLAC normalization + smarter sample rate handling (no more weird pitch shifts!).

✅ Install improvements: Docs and setup scripts updated for smoother cross-platform installs (Linux/macOS/Windows).

✅ Bug squashes: Fixed edge cases with special chars, long texts, and audio truncation — making samples more reliable.

📦 Bonus: Docker builds and precompiled binaries may be included (check the release assets!).

🔗 Grab it here: v3.2.0 Release

🛠️ Need help spinning up a custom wake word dataset? Just ask — happy to walk through examples! 🧠✨

🔗 View Release

Ollama – v0.17.8-rc2: mlx: perf improvements (#14768)

🚀 Ollama v0.17.8-rc2 is here — and it’s bringing major mlx performance boosts for Apple Silicon users!

This release is all about speed and efficiency on M1/M2/M3 chips, thanks to smarter use of Apple’s MLX framework. Here’s what’s new:

🔹 Layer Norm Got a Power-Up

→ Ditched the 6-step manual layer norm (mean → subtract → variance → rsqrt → multiply → add)

→ Now uses `mlx_fast_layer_norm` — a native, optimized kernel. Way faster and cleaner!

🔹 GQA Just Got Smarter

→ Removed custom `RepeatKV` tiling logic for Grouped-Query Attention (GQA)

→ Now leverages `scaled_dot_product_attention`, which natively supports GQA — as long as `n_q_heads % n_kv_heads == 0`.

✅ Result?

⚡ Faster inference

🧠 Lower memory usage

✨ Cleaner, more maintainable code

Perfect for devs and tinkerers pushing their Macs to the limit! 🍏💻

Let us know if you’d like a deep dive into how GQA + native attention works under the hood! 🧠⚡

🔗 View Release

Deep-Live-Cam – 2.7 beta

🚨 AI Enthusiasts—deepfake magic just got a serious upgrade! 🚨

🔥 DeepLiveCam v2.7 Beta is here—celebrating 80K GitHub stars with the biggest update yet! 🎉

✨ What’s Fresh:

✅ Realtime Face Enhancer — now running up to 27 FPS with enhancement on! 🚀

✅ Inswapper Optimizer — faster, smoother swaps with less lag

✅ 2 NEW enhancer models: GPEN 512 & GPEN 256 (for sharper, more natural results)

✅ Face Enhancer Scaler — dial in enhancement intensity like a pro 🎛️

✅ Masking upgrades: Quick Lip, Lip, Chin, Eyes masks — precision control at your fingertips 🎭

✅ Interpolation support for buttery-smooth output frames

✅ GPU Changer — full multi-GPU support (switch GPUs on the fly!) 🖥️➡️🖥️

✅ LUTs — apply color grading presets in seconds 🎨

✅ Window Projection — watch live output in a dedicated (even full-screen!) window 📺

✅ Camera Refresh & Resolution Changer — tweak input/output like a streamer

✅ Cleaner, smarter UI — because UX matters as much as power 💡

🎥 Check out the demo: `2026-03-12.02.34.57.mp4`

📚 Quickstart guide: [link in bio/pinned]

This isn’t just an update—it’s a deepfake renaissance. Go play! 🧪✨

🔗 View Release

Voxtral Wyoming – v2.1.0

🚨 Voxtral Wyoming v2.1.0 is live! 🚨

Offline STT just got smarter and more flexible—and you’re going to love these updates:

🔹 ✨ Word Replacement Feature

Swap words or phrases on the fly in your prompts—ideal for A/B testing, localization, brand tone adjustments, or just having fun with custom phrasing. Think of it as prompt surgery 🩺: precise, fast, and fully in your control.

🔹 🚀 Early Request Acceptance

Want a feature before it drops publicly? Submit your request early and lock in priority access—get ahead of the curve and shape what’s next.

💡 Bonus: Still supports CPU/CUDA/MPS, Dockerized deployment, and auto-converts MP3/OGG/FLAC/WAV → PCM16.

🔧 Configurable via env vars for host, port, language, model ID—and now word replacement rules too.

Ready to give it a spin? Drop your favorite use case below—we’re listening! 🎙️🤖

🔗 View Release