Models
80 open-weight models, sized
Updated 8 Oct 2026 1 new this month
Parameter counts, layer counts and attention shapes are read from each model's own config, because the KV-cache arithmetic depends on them exactly. Sizes below are at Q4_K_M and 8K context against a Radeon RX 7900 XTX. Missing a model that shipped this week? Open an issue — the table is hand-maintained.
| Model | Released | Params | Quant | Weights | Max context | Licence | On this rig | |
|---|---|---|---|---|---|---|---|---|
| Qwen3 0.6B Useful mostly as a speculative-decoding draft model for its larger siblings. |
Apr 2025 | 0.6B | Q4_K_M | 0.3 GB | 32K | Apache 2.0 | 1723 tok/s | Check · Hardware |
| Qwen3.5 0.8B Draft model for speculative decoding, or a classifier that fits in 1 GB. |
28 Feb 2026 | 0.87B | Q4_K_M | 0.5 GB | 256K | Apache 2.0 | 1188 tok/s | Check · Hardware |
| Gemma 3 1B Text-only. Single KV head makes its cache almost free at long context. |
Mar 2025 | 1.0B | Q4_K_M | 0.6 GB | 32K | Gemma Terms of Use | 1034 tok/s | Check · Hardware |
| Llama 3.2 1B Instruct The smallest Llama worth running. Fits anywhere, including phones and 4 GB cards. |
Sep 2024 | 1.24B | Q4_K_M | 0.7 GB | 128K | Llama 3.2 Community | 834 tok/s | Check · Hardware |
| Qwen3 1.7B Punches above its size on structured tasks, with optional thinking mode. |
Apr 2025 | 1.72B | Q4_K_M | 1.0 GB | 32K | Apache 2.0 | 601 tok/s | Check · Hardware |
| Qwen3.5 2B Phone-class, and multimodal. Replaces Llama 3.2 3B as the "it runs on anything" answer. |
28 Feb 2026 | 2.27B | Q4_K_M | 1.3 GB | 256K | Apache 2.0 | 455 tok/s | Check · Hardware |
| LFM2.5 2.6B Convolution-heavy hybrid for CPUs and NPUs: 22 of 30 blocks keep no KV cache at all. |
28 Jul 2026 | 2.7B | Q4_K_M | 1.5 GB | 128K | LFM Open License v1.0 | 383 tok/s | Check · Hardware |
| Llama 3.2 3B Instruct The 2024 "it just runs" model for 8 GB laptops. Qwen3.5 4B does the same job better now. |
Sep 2024 | 3.21B | Q4_K_M | 1.8 GB | 128K | Llama 3.2 Community | 322 tok/s | Check · Hardware |
| Granite 4.1 3B Dense, small, enterprise-flavoured: tool calling and instruction following, no thinking mode. |
29 Apr 2026 | 3.4B | Q4_K_M | 1.9 GB | 128K | Apache 2.0 | 304 tok/s | Check · Hardware |
| Phi-4-mini 3.8B MIT-licensed, dense, and unusually strong on instruction following for its size. |
Feb 2025 | 3.84B | Q4_K_M | 2.2 GB | 128K | MIT | 269 tok/s | Check · Hardware |
| Ministral 3 3B Edge model with a vision encoder and a 256K window. Apache 2.0. |
Dec 2025 | 3.85B | Q4_K_M | 2.2 GB | 256K | Apache 2.0 | 268 tok/s | Check · Hardware |
| Qwen3 4B The 2025 sweet spot for 8 GB cards with reasoning traces. Qwen3.5 4B adds vision and 8× the context. |
Apr 2025 | 4.02B | Q4_K_M | 2.3 GB | 32K | Apache 2.0 | 257 tok/s | Check · Hardware |
| Gemma 3 4B Vision-capable at 4B. Superseded by Gemma 4 E4B, still everywhere. |
Mar 2025 | 4.3B | Q4_K_M | 2.4 GB | 128K | Gemma Terms of Use | 240 tok/s | Check · Hardware |
| Qwen3.5 4B The 8 GB coding agent. Q4 lands near 3.4 GB, leaving room for a real context window. |
28 Feb 2026 | 4.66B | Q4_K_M | 2.6 GB | 256K | Apache 2.0 | 222 tok/s | Check · Hardware |
| Gemma 4 E2B "E2B" is 2.3B effective, but the file holds 5B because of per-layer embeddings — size it as 5B. Text, image and audio in. |
2 Apr 2026 | 5.1B | Q4_K_M | 2.9 GB | 128K | Apache 2.0 | 203 tok/s | Check · Hardware |
| Mistral 7B Instruct v0.3 Old but extremely well behaved, and permissively licensed for commercial use. |
May 2024 | 7.25B | Q4_K_M | 4.1 GB | 32K | Apache 2.0 | 143 tok/s | Check · Hardware |
| Olmo 3 7B Instruct Fully open — training data and code included. Full multi-head attention, so its cache is 4× a GQA 7B at the same context. |
20 Nov 2025 | 7.3B | Q4_K_M | 4.1 GB | 64K | Apache 2.0 | 142 tok/s | Check · Hardware |
| DeepSeek-R1-Distill-Qwen 7B Reasoning traces on a 7B budget. Expect long outputs — budget context accordingly. |
Jan 2025 | 7.62B | Q4_K_M | 4.3 GB | 128K | MIT | 136 tok/s | Check · Hardware |
| Qwen2.5-Coder 7B The standard local autocomplete model — small enough to keep resident all day, and still the best FIM model under 8B. |
Nov 2024 | 7.62B | Q4_K_M | 4.3 GB | 128K | Apache 2.0 | 136 tok/s | Check · Hardware |
| Ling 3.0 Tiny 7.9B-A1.3B An 8B MoE with 1.3B active and a latent KV cache on only 6 of 24 layers — reasoning and tool use sized for Apple Silicon and edge boxes. |
10 Aug 2026 | 7.9B MoE | Q4_K_M | 4.4 GB | 128K | MIT | 306 tok/s | Check · Hardware |
| Gemma 4 E4B The laptop Gemma. 4.5B effective, 8B on disk; a single KV head per window layer keeps its cache tiny. |
2 Apr 2026 | 8.0B | Q4_K_M | 4.5 GB | 128K | Apache 2.0 | 129 tok/s | Check · Hardware |
| Llama 3.1 8B Instruct Still the most widely deployed local model, with the largest fine-tune ecosystem. Not the strongest 8B any more. |
Jul 2024 | 8.03B | Q4_K_M | 4.5 GB | 128K | Llama 3.1 Community | 129 tok/s | Check · Hardware |
| Qwen3 8B Apache-2.0 alternative to Llama 3.1 8B, with a switchable thinking mode. |
Apr 2025 | 8.19B | Q4_K_M | 4.6 GB | 128K | Apache 2.0 | 126 tok/s | Check · Hardware |
| Fara 7B A web computer-use agent on a Qwen2.5-VL base — it clicks, fills forms and stops for permission. Not a chat model; size it like an 8B with vision. |
24 Nov 2025 | 8.29B | Q4_K_M | 4.7 GB | 125K | MIT | 125 tok/s | Check · Hardware |
| LFM2.5 8B-A1B An 8B MoE with ~1.5B active, aimed at laptops without a GPU. Licence is permissive below $10M revenue. |
28 May 2026 | 8.47B MoE | Q4_K_M | 4.8 GB | 125K | LFM Open License v1.0 | 265 tok/s | Check · Hardware |
| Granite 4.1 8B Matches the old Granite 4.0 32B MoE at a quarter of the size. Fast, Apache 2.0, no reasoning traces. |
29 Apr 2026 | 8.79B | Q4_K_M | 4.9 GB | 128K | Apache 2.0 | 118 tok/s | Check · Hardware |
| Ministral 3 8B Mistral's 8B with images in. Plain GQA, so budget more KV cache than Qwen3.5 9B at the same context. |
Dec 2025 | 8.92B | Q4_K_M | 5.0 GB | 256K | Apache 2.0 | 116 tok/s | Check · Hardware |
| Ornith 1.5 9B The small Ornith: a coding-agent reasoning build on the Qwen3.5 9B architecture. Same VRAM as its base, thinks before every answer. |
19 Aug 2026 | 9.41B | Q4_K_M | 5.3 GB | 256K | MIT | 110 tok/s | Check · Hardware |
| Qwen3.5 9B The default for 8–12 GB cards in 2026: beats every older 8B on every published benchmark, with vision. |
28 Feb 2026 | 9.65B | Q4_K_M | 5.4 GB | 256K | Apache 2.0 | 107 tok/s | Check · Hardware |
| Gemma 4 12B The "unified" Gemma 4: text, image and audio in one 12B that fits a 12 GB card at Q4. 140+ languages. |
29 May 2026 | 12B | Q4_K_M | 6.7 GB | 256K | Apache 2.0 | 86 tok/s | Check · Hardware |
| Gemma 3 12B Strong multilingual chat with images, sized for 12–16 GB cards. Gemma 4 12B is the same size and better. |
Mar 2025 | 12.2B | Q4_K_M | 6.9 GB | 128K | Gemma Terms of Use | 85 tok/s | Check · Hardware |
| Mistral NeMo 12B Multilingual 12B with a 128K window, built with NVIDIA. A roleplay and fiction favourite that refuses to die. |
Jul 2024 | 12.2B | Q4_K_M | 6.9 GB | 128K | Apache 2.0 | 85 tok/s | Check · Hardware |
| Ministral 3 14B The largest Ministral. A 12 GB card runs it at Q4 with a few gigabytes to spare. |
Dec 2025 | 13.9B | Q4_K_M | 7.8 GB | 256K | Apache 2.0 | 74 tok/s | Check · Hardware |
| Phi-4 14B Trained heavily on synthetic reasoning data. Short 16K window is its main limitation. |
Dec 2024 | 14.7B | Q4_K_M | 8.3 GB | 16K | MIT | 70 tok/s | Check · Hardware |
| Qwen3 14B The largest Qwen3 that fits a 12 GB card at Q4 with room for context. |
Apr 2025 | 14.8B | Q4_K_M | 8.3 GB | 128K | Apache 2.0 | 70 tok/s | Check · Hardware |
| DeepSeek-R1-Distill-Qwen 14B The 2025 reasoning-per-gigabyte pick for a 12 GB card. Qwen3.5 9B in thinking mode has since overtaken it. |
Jan 2025 | 14.8B | Q4_K_M | 8.3 GB | 128K | MIT | 70 tok/s | Check · Hardware |
| Qwen2.5-Coder 14B Noticeably better at whole-file edits than the 7B, still comfortable on 12 GB. |
Nov 2024 | 14.8B | Q4_K_M | 8.3 GB | 128K | Apache 2.0 | 70 tok/s | Check · Hardware |
| gpt-oss 20B Ships natively in MXFP4, so the 4-bit weights are the reference weights, not a lossy copy. Fits 16 GB. |
Aug 2025 | 20.9B MoE | MXFP4 | 10.8 GB | 128K | Apache 2.0 | 120 tok/s | Check · Hardware |
| Mistral Small 3.2 24B Apache-2.0, vision-capable, and the most 24 GB-friendly of the 2025 generalists. |
Jun 2025 | 23.6B | Q4_K_M | 13.3 GB | 128K | Apache 2.0 | 44 tok/s | Check · Hardware |
| Devstral Small 2 24B Built for software-engineering agents (OpenHands, Cline). Dense 24B, 384K window. Not in the Ollama library. |
Dec 2025 | 24B | Q4_K_M | 13.5 GB | 384K | Apache 2.0 | 43 tok/s | Check · Hardware |
| Gemma 4 26B-A4B Mixture of experts with 3.8B active. Slower to think than Qwen3.6 35B-A3B, faster to answer, and it sees images. |
2 Apr 2026 | 26.5B MoE | Q4_K_M | 14.9 GB | 256K | Apache 2.0 | 105 tok/s | Check · Hardware |
| Gemma 3 27B The 2025 single-GPU generalist with vision. Its Gemma-licence terms are the reason to prefer Gemma 4 now. |
Mar 2025 | 27.4B | Q4_K_M | 15.4 GB | 128K | Gemma Terms of Use | 38 tok/s | Check · Hardware |
| Qwen3.8 27B The current default local Qwen: dense 27B, text + image + video, 262K context. Only 16 of its 64 blocks keep a KV cache, so long context is cheap. |
14 Aug 2026 | 27.8B | Q4_K_M | 15.6 GB | 256K | Apache 2.0 | 37 tok/s | Check · Hardware |
| Qwen3.6 27B The 24 GB coding pick of spring 2026 (77.2 SWE-bench Verified). Same shape as 3.8, one generation behind. |
22 Apr 2026 | 27.8B | Q4_K_M | 15.6 GB | 256K | Apache 2.0 | 37 tok/s | Check · Hardware |
| Granite 4.1 30B The largest Granite. Dense 29B at Q4 is a comfortable 24 GB fit. |
29 Apr 2026 | 28.9B | Q4_K_M | 16.3 GB | 128K | Apache 2.0 | 36 tok/s | Check · Hardware |
| Muse Glimmer 30B Meta's first open weights since Llama 4: a dense 30B distilled from Muse Spark for always-on local agents. Two KV heads keep the cache small. |
10 Aug 2026 | 29.8B | Q4_K_M | 16.8 GB | 128K | Apache 2.0 | 35 tok/s | Check · Hardware |
| Qwen3 30B-A3B The MoE that made "3B active" a category. Qwen3.6 35B-A3B is its direct replacement. |
Apr 2025 | 30.5B MoE | Q4_K_M | 17.1 GB | 128K | Apache 2.0 | 120 tok/s | Check · Hardware |
| Qwen3 Coder 30B-A3B Agentic coding MoE with a 256K native window. Still the most-downloaded local code model. |
Jul 2025 | 30.5B MoE | Q4_K_M | 17.1 GB | 256K | Apache 2.0 | 120 tok/s | Check · Hardware |
| GLM-4.7-Flash 30B-A3B MIT-licensed 30B-A3B tuned for agentic coding, with a DeepSeek-style latent KV cache. 60–80 tok/s reported on a 4090. |
20 Jan 2026 | 31.2B MoE | Q4_K_M | 17.5 GB | 198K | MIT | 133 tok/s | Check · Hardware |
| Gemma 4 31B The dense flagship: strongest maths of the 24–32 GB class (89% AIME), clean prose, vision. Q4 is a tight 24 GB fit. |
2 Apr 2026 | 31.3B | Q4_K_M | 17.6 GB | 256K | Apache 2.0 | 33 tok/s | Check · Hardware |
| Nemotron 3.5 Lightning 30B-A3B Mamba-2 + MoE hybrid built for the execution layer of agents: only 6 attention blocks, so the KV cache is almost free. Weights, data and recipe all open. |
11 Aug 2026 | 31.6B MoE | Q4_K_M | 17.8 GB | 256K | OpenMDW-1.1 | 124 tok/s | Check · Hardware |
| Olmo 3.1 32B Instruct The largest fully open model you can audit end to end. Q4 fits 24 GB, tightly. |
10 Dec 2025 | 32.2B | Q4_K_M | 18.1 GB | 64K | Apache 2.0 | 32 tok/s | Check · Hardware |
| Qwen3 32B The classic 24 GB target, and still the strongest local translator under 70B. Qwen3.8 27B is smaller and better at everything else. |
Apr 2025 | 32.8B | Q4_K_M | 18.4 GB | 128K | Apache 2.0 | 32 tok/s | Check · Hardware |
| DeepSeek-R1-Distill-Qwen 32B MIT-licensed and close to the 70B distill on maths. A 24 GB card handles it at Q4. |
Jan 2025 | 32.8B | Q4_K_M | 18.4 GB | 128K | MIT | 32 tok/s | Check · Hardware |
| Qwen2.5-Coder 32B The first local code model that felt competitive with hosted assistants. Qwen3.8 27B is smaller and far ahead on agentic work. |
Nov 2024 | 32.8B | Q4_K_M | 18.4 GB | 128K | Apache 2.0 | 32 tok/s | Check · Hardware |
| LLM-jp 4 33B Thinking Japan’s national-institute reasoning model, Japanese and English. A plain dense Llama-style 33B: Q4 is a tight 24 GB fit. |
14 Aug 2026 | 33.2B | Q4_K_M | 18.7 GB | 64K | Apache 2.0 | 31 tok/s | Check · Hardware |
| Qwen3.6 35B-A3B Mixture of experts with ~3B active: the fastest serious model a 24 GB card runs, and the best MoE under 40B on agentic coding. |
16 Apr 2026 | 35.9B MoE | Q4_K_M | 20.2 GB | 256K | Apache 2.0 | 120 tok/s | Check · Hardware |
| Ornith 1.5 35B-A3B A reasoning-first MIT build on the Qwen3.6 35B-A3B architecture (thinks before every answer). Same VRAM as its base. |
19 Aug 2026 | 35.9B MoE | Q4_K_M | 20.2 GB | 256K | MIT | 120 tok/s | Check · Hardware |
| Llama 3.3 70B Instruct Still the creative-writing favourite: consistent voice, takes direction. Needs 48 GB to sit comfortably on GPU at Q4. |
Dec 2024 | 70.6B | Q4_K_M | 39.7 GB | 128K | Llama 3.3 Community | offload | Check · Hardware |
| DeepSeek-R1-Distill-Llama 70B The strongest of the R1 distills, and the one that most needs 48 GB or more. |
Jan 2025 | 70.6B | Q4_K_M | 39.7 GB | 128K | MIT / Llama 3.3 Community | offload | Check · Hardware |
| Llama 4 Scout 109B-A17B A 10M-token window on paper, chunked attention in practice (8K chunks on 3 of 4 layers). Needs 64 GB+ at Q4. |
5 Apr 2025 | 109B MoE | Q4_K_M | 61.3 GB | 1024K | Llama 4 Community | won't fit | Check · Hardware |
| gpt-oss 120B Designed to land on one 80 GB card. Only ~5B parameters are active per token. |
Aug 2025 | 117B MoE | MXFP4 | 60.5 GB | 128K | Apache 2.0 | won't fit | Check · Hardware |
| Mistral Small 4 119B-A6B Instruct, reasoning, vision and code in one 119B MoE. A latent KV cache (320 wide) keeps context cheap; the weights still want 64 GB+. |
17 Mar 2026 | 119B MoE | Q4_K_M | 66.9 GB | 256K | Apache 2.0 | won't fit | Check · Hardware |
| Nemotron 3 Super 120B-A12B The open-training-data 120B. Same hybrid layout as Lightning, so 128K context costs under a gigabyte. |
Mar 2026 | 124B MoE | Q4_K_M | 69.7 GB | 256K | NVIDIA Open Model | won't fit | Check · Hardware |
| Ling 3.0 Flash 124B-A5B A 124B hybrid (5 linear-attention layers per MLA layer) with 5.1B active: SWE-bench Pro 56.6 and AIME 93 claimed. Built for 96–128 GB machines. |
2 Aug 2026 | 124B MoE | Q4_K_M | 69.7 GB | 256K | MIT | won't fit | Check · Hardware |
| Qwen3.5 122B-A10B The 96–128 GB unified-memory model: 122B of knowledge at 10B-active speed. |
24 Feb 2026 | 125B MoE | Q4_K_M | 70.3 GB | 256K | Apache 2.0 | won't fit | Check · Hardware |
| Qwen3.8-Flash-Next 180B-A6B The open preview of the Qwen4 architecture: a 125B-A6B hybrid (Gated DeltaNet + sparse attention, KV cache on 12 of 48 blocks) plus a 51B n-gram embedding and a 4B draft head — 180B on disk, 6B active. The hosted "Qwen3.8-Flash" is this model with a 1M window. |
24 Aug 2026 | 180B MoE | Q4_K_M | 101.2 GB | 256K | Qwen Community 1.0 | won't fit | Check · Hardware |
| Qwen3 235B-A22B Workstation class. Realistically a 192 GB unified-memory or multi-GPU model. |
Apr 2025 | 235B MoE | Q4_K_M | 132.1 GB | 128K | Apache 2.0 | won't fit | Check · Hardware |
| DeepSeek V4 Flash 284B-A13B The V4 for 192–256 GB machines: Q3 squeezes into 192 GB, Q4 wants 256 GB, and 128 GB falls short even at Q2. Cache is modelled as a 576-wide latent; V4 compresses it further at long context, so this is conservative. |
31 Jul 2026 | 284B MoE | Q4_K_M | 159.7 GB | 1024K | MIT | won't fit | Check · Hardware |
| GLM-5.3-Flash 320B-A18B The first natively multimodal GLM-5 and the first hybrid: 34 linear-attention blocks and 11 sparse-attention blocks with a 512-wide latent cache, so a 1M window stays affordable. Z.ai says it beats GLM-5.2 at 18B active; MIT. |
25 Aug 2026 | 321B MoE | Q4_K_M | 180.5 GB | 1024K | MIT | won't fit | Check · Hardware |
| Ornith 1.5 397B-A17B The flagship Ornith on the Qwen3.5 397B-A17B architecture, MIT-licensed. A 256 GB Mac Studio at Q4, and it is in the Ollama library. |
19 Aug 2026 | 397B MoE | Q4_K_M | 223.2 GB | 256K | MIT | won't fit | Check · Hardware |
| Qwen3.5 397B-A17B Flagship-class at 17B active. A 256 GB Mac Studio or a multi-GPU box at Q4. |
16 Feb 2026 | 403B MoE | Q4_K_M | 226.6 GB | 256K | Apache 2.0 | won't fit | Check · Hardware |
| DeepSeek V4.1 Flash 552B-A16B New Twice V4 Flash's backbone (552B, 16B active per generated token) plus 196B of Engram lookup tables (189 GiB at FP8) that DwarfStar streams from SSD, so they are not counted here. A 512 GB Mac model, and as of October 2026 no mainline llama.cpp or Ollama build: DwarfStar on a Mac, vLLM across four GPUs. DeepSeek puts the cache at 890 bytes/token; modelled conservatively as 4 latent layers plus a 128-token window. |
10 Sep 2026 | 552B MoE | Q4_K_M | 310.4 GB | 1024K | MIT | won't fit | Check · Hardware |
| DeepSeek-R1 671B The January 2025 moment. Multi-head latent attention keeps its KV cache tiny; the weights do not. |
Jan 2025 | 671B MoE | Q4_K_M | 377.3 GB | 128K | MIT | won't fit | Check · Hardware |
| GLM-5.2 744B-A40B The strongest all-round open-weight model of mid-2026 on most public boards. Listed as a ceiling: 512 GB of unified memory at Q4. |
13 Jun 2026 | 753B MoE | Q4_K_M | 423.4 GB | 1024K | MIT | won't fit | Check · Hardware |
| GLM-5.3 744B-A40B Same base as GLM-5.2, new post-training: Z.ai's most capable open-weights coder (Terminal-Bench 3.0 28.3, DeepSWE 66.9). Not MIT any more — its own GLM-5.3 licence. Same ceiling as 5.2: 512 GB of unified memory at Q4. |
25 Aug 2026 | 753B MoE | Q4_K_M | 423.4 GB | 1024K | GLM-5.3 License | won't fit | Check · Hardware |
| Kimi K2.6 1T-A32B The open coding-agent benchmark leader of spring 2026 (80.2 SWE-bench). A 512 GB Mac Studio pair, or a ceiling. |
20 Apr 2026 | 1027B MoE | Q4_K_M | 577.5 GB | 256K | Modified MIT | won't fit | Check · Hardware |
| DeepSeek V4 Pro 1.6T-A49B The 0813 refresh of the V4 flagship. Included as the honest ceiling; a terabyte of weights at Q4. |
13 Aug 2026 | 1650B MoE | Q4_K_M | 927.8 GB | 1024K | MIT | won't fit | Check · Hardware |
| Qwen3.8 2.4T-A95B The first open Qwen-Max-class flagship. Listed as the honest ceiling: nothing short of a rack runs it. |
12 Aug 2026 | 2446B MoE | Q4_K_M | 1375.4 GB | 256K | Qwen3.8-Max License | won't fit | Check · Hardware |
| Kimi K3 2.8T-A104B The largest open-weight model ever published. 69 of 93 blocks are linear attention, so its cache is tiny; its 2.8T weights are the problem. |
27 Jul 2026 | 2780B MoE | Q4_K_M | 1563.2 GB | 1024K | Modified MIT | won't fit | Check · Hardware |
Weight sizes are computed from the parameter count and the quantisation's effective bits per weight, not read off a file listing — expect them to land within a few percent of the GGUF you actually download. The method, in full.