https://quantized.uk/ New models, data cadence, and site updates for LLM quantization intelligence. en Sat, 08 Aug 2026 12:00:00 GMT https://quantized.uk/#changelog changelog-2026-08-08-New cookbook: running GPT-OSS 20B/120B l Sat, 08 Aug 2026 12:00:00 GMT New cookbook: running GPT-OSS 20B/120B locally without re-quantizing (23 guides). VRAM calculator now offers MXFP4 — picking Q4_K_M for GPT-OSS overstated weights by ~14% https://quantized.uk/#changelog changelog-2026-08-07-Model index +4 → 75: GPT-OSS 20B/120B (n Fri, 07 Aug 2026 12:00:00 GMT Model index +4 → 75: GPT-OSS 20B/120B (native MXFP4), GLM-4.5-Air 106B-A12B, Devstral Small 1.1 — MoE-heavy batch for 16GB cards and unified-memory Macs https://quantized.uk/#changelog changelog-2026-07-22-Polish: re-rendered og.png (71+ models), Wed, 22 Jul 2026 12:00:00 GMT Polish: re-rendered og.png (71+ models), all 22 cookbook guides have verified stack banners https://quantized.uk/#changelog changelog-2026-07-22-Cadence pack: +4 models (Gemma 3 27B, R1 Wed, 22 Jul 2026 12:00:00 GMT Cadence pack: +4 models (Gemma 3 27B, R1-Llama-8B, Phi-4, Qwen3 1.7B), superseded tags, measured/estimated labels, Hub “recent”, weekly updates, RSS, cookbook verified stack (71 models) https://quantized.uk/#changelog changelog-2026-06-26-UX for real traffic: job paths, mobile G Fri, 26 Jun 2026 12:00:00 GMT UX for real traffic: job paths, mobile GPU profile, OG/favicon, honest format heat, feedback email https://quantized.uk/#changelog changelog-2026-06-26-Model index +4: Qwen3 4B, Qwen3-Coder 30 Fri, 26 Jun 2026 12:00:00 GMT Model index +4: Qwen3 4B, Qwen3-Coder 30B-A3B, Mistral Large 3, GLM-4-9B (67 total) https://quantized.uk/#changelog changelog-2026-06-26-QA fixes: HF stats merge on failure, ≤3B Fri, 26 Jun 2026 12:00:00 GMT QA fixes: HF stats merge on failure, ≤3B filter, CLI/VRAM tool bugs, i18n polish https://quantized.uk/#changelog changelog-2026-06-26-Model index +5: Qwen3 32B, 30B-A3B MoE, Fri, 26 Jun 2026 12:00:00 GMT Model index +5: Qwen3 32B, 30B-A3B MoE, 235B-A22B, DeepSeek-V3, DeepSeek-R1 (63 total) https://quantized.uk/#changelog changelog-2026-06-26-Model index +7: Qwen3 8B/14B, Gemma 3 4B Fri, 26 Jun 2026 12:00:00 GMT Model index +7: Qwen3 8B/14B, Gemma 3 4B/12B, Llama 4 Scout/Maverick, Llama 3.1 405B (58 total) https://quantized.uk/#changelog changelog-2026-06-25-Cookbook TOC scroll highlight, code bloc Thu, 25 Jun 2026 12:00:00 GMT Cookbook TOC scroll highlight, code block copy, Quant Hub Markdown export https://quantized.uk/#changelog changelog-2026-06-25-Cookbook reading progress bar, model HF Thu, 25 Jun 2026 12:00:00 GMT Cookbook reading progress bar, model HF link copy, Quant Hub shareable filter URLs https://quantized.uk/#changelog changelog-2026-06-25-Breadcrumb nav + JSON-LD, cookbook artic Thu, 25 Jun 2026 12:00:00 GMT Breadcrumb nav + JSON-LD, cookbook article TOC, Quant Hub GPU quick-filter chips https://quantized.uk/quant-hub/qwen3-4b/ https://quantized.uk/quant-hub/qwen3-4b/ Fri, 26 Jun 2026 12:00:00 GMT Smallest Qwen3 dense with thinking mode. Q4 ~3.2GB — ideal for 8GB GPUs and edge devices. https://quantized.uk/quant-hub/qwen3-coder-30b-a3b/ https://quantized.uk/quant-hub/qwen3-coder-30b-a3b/ Fri, 26 Jun 2026 12:00:00 GMT Agentic coding MoE with 3.3B active params and 256K native context. Top open coder for 16–24GB cards. https://quantized.uk/quant-hub/mistral-large-3/ https://quantized.uk/quant-hub/mistral-large-3/ Fri, 26 Jun 2026 12:00:00 GMT Mistral 3 flagship MoE (41B active / 675B total) with vision encoder. FP8 on 8×H200; GGUF quant for research clusters only. https://quantized.uk/quant-hub/glm-4-9b-chat/ https://quantized.uk/quant-hub/glm-4-9b-chat/ Fri, 26 Jun 2026 12:00:00 GMT Zhipu GLM-4 open 9B with 128K context, tool calling, and strong bilingual (EN/ZH) performance. https://quantized.uk/quant-hub/gemma-3-27b-it/ https://quantized.uk/quant-hub/gemma-3-27b-it/ Wed, 22 Jul 2026 12:00:00 GMT Gemma 3 large instruct with long context and multimodal support. Q4 ~16GB — dual-GPU or 24GB card with short ctx. https://quantized.uk/quant-hub/deepseek-r1-distill-llama-8b/ https://quantized.uk/quant-hub/deepseek-r1-distill-llama-8b/ Wed, 22 Jul 2026 12:00:00 GMT R1 reasoning distilled into Llama 3.1 8B. Best chain-of-thought for 8–12GB cards; huge community GGUF support. https://quantized.uk/quant-hub/phi-4/ https://quantized.uk/quant-hub/phi-4/ Wed, 22 Jul 2026 12:00:00 GMT Microsoft Phi-4 dense 14B — strong reasoning for size. Q4 ~9GB fits 12GB cards with moderate context. https://quantized.uk/quant-hub/qwen3-1.7b/ https://quantized.uk/quant-hub/qwen3-1.7b/ Wed, 22 Jul 2026 12:00:00 GMT Tiny Qwen3 with thinking mode. Q4 ~1.4GB — phones, NUC, and always-on local agents. https://quantized.uk/quant-hub/gpt-oss-20b/ https://quantized.uk/quant-hub/gpt-oss-20b/ Fri, 07 Aug 2026 12:00:00 GMT OpenAI open-weight MoE (21B total / 3.6B active), shipped natively in MXFP4 — ~12.8GB runs on a 16GB card with no quality tax. Only 3.6B active params means CPU-offload stays usable. https://quantized.uk/quant-hub/gpt-oss-120b/ https://quantized.uk/quant-hub/gpt-oss-120b/ Fri, 07 Aug 2026 12:00:00 GMT The big GPT-OSS (117B total / 5.1B active). Native MXFP4 checkpoint is ~61GB — fits one 80GB card or a 128GB unified-memory Mac. Partial offload on 24GB consumer cards is slow but works. https://quantized.uk/quant-hub/glm-4.5-air/ https://quantized.uk/quant-hub/glm-4.5-air/ Fri, 07 Aug 2026 12:00:00 GMT Zhipu's agentic/reasoning MoE (106B total / 12B active). Q4 ~64GB — the sweet spot is a 96GB+ unified Mac or 2× 48GB cards. Strong tool-calling for its class. https://quantized.uk/quant-hub/devstral-small-2507/ https://quantized.uk/quant-hub/devstral-small-2507/ Fri, 07 Aug 2026 12:00:00 GMT Mistral + All Hands agentic coding model on a Mistral Small 3.1 base. Built for repo-scale tool use rather than single-file completion. Q4 ~14GB fits a 16GB card.