Unsloth shipped v0.1.900-beta on 2026-09-28 (PyPI unsloth 2026.9.12): Desktop runs Laya Decision models locally via a Jev-compatible /v1/systemone endpoint, adds a Skills CRUD editor and Library viewers for PDF/Office, and improves Apple Silicon with batched serving, structured outputs, and TurboQuant KV. Media pipelines claim ~4.5× LTX-2.3 clips, 1.7–6.3× VAE decode, and up to ~1 minute faster MiniMax-H3 first render with 25–29 GiB lower peak VRAM.
Key Takeaways
- ✓Shipped v0.1.900-beta — local Laya Decision API + Skills editor + Library viewers
- ✓LTX-2.3 ~4.5×; VAE decode 1.7–6.3×; MiniMax-H3 first render up to ~1 min faster, −25–29 GiB peak VRAM
- ✓Apple Silicon: batched serving, structured outputs, TurboQuant KV; ModelScope downloads
- ✓Docs: docs.unsloth.ai
Key Decision Metrics at a Glance
Turn your technical choice into a development budget
Compare 40 dev plans & simulate token costs vs $20/mo subscriptions
Project Links & Resources
Direct AccessIn-Depth Technical Analysis
Core Background & Industry Pain Points
Local fine-tuning/Studio stacks must cover decision micro-models, Agent Skills, and multimodal generation. Gaps: Jev/TypeSafe often cloud-bound; Skills edited outside the app; Office/PDF previews missing; Apple Silicon lagging CUDA on batching/KV quant; video first-render and VAE decode still dominate latency/VRAM.
Architecture Highlights & Internals
v0.1.900-beta adds a Desktop Decision API for local Laya models via Jev-compatible /v1/systemone (MLX on Apple Silicon, #11603), Skills CRUD, Library PDF/Office viewers, ModelScope downloads, batched serving, grammar-constrained response_format, and TurboQuant KV. Training gains faster block-FP8 LoRA and LoRA-on-W8A8 compressed tensors; FastModel loads Voxtral/Qwen2-Audio.
Authoritative Benchmarks & Measured Scores
Vendor-reported: LTX-2.3 clips ~4.5×; VAE decode 1.7–6.3×; MiniMax-H3 first render up to ~1 minute faster with 25–29 GiB lower peak VRAM on A100/B200/RTX PRO 6000. No third-party SWE-bench table—reproduce on your GPU/MLX before capacity planning.
Developer Hands-on Guide
Upgrade to v0.1.900-beta or pip install -U unsloth (PyPI 2026.9.12). Enable Decision API under Settings > API; manage Skills in-app; preview docs in Library. See docs.unsloth.ai. Beta tag—pin and regression-test before prod.
Benchmark side-by-side against alternatives, or calculate monthly token cost vs subscription break-even.
Discussion & Comments
0Sign in to join the discussion
Connect with AI developers to exchange benchmark insights.