{"name":"Twoody local LLM calculator data","version":"2026-09-26.1","reviewed_on":"2026-09-26","license":"CC BY 4.0","license_url":"https://creativecommons.org/licenses/by/4.0/","url":"https://www.twoody.com/local-llm-calculator","method":{"reference_model":"LLaMA 2 7B Q4_0 (3.8226 GB)","gpu_efficiency":0.85,"cpu_efficiency":0.55,"mac_gpu_share":{"up to 32 GB":0.667,"36 to 128 GB":0.75,"above 128 GB":0.8},"document_tokens":14000},"sources":{"apple":"https://www.apple.com/newsroom/","apple_specs":"https://support.apple.com/specs","llama_cpp_apple":"https://github.com/ggml-org/llama.cpp/discussions/4167","llama_cpp_cuda":"https://github.com/ggml-org/llama.cpp/discussions/15013","llama_cpp_vulkan":"https://github.com/ggml-org/llama.cpp/discussions/10879","llama_cpp_gpt_oss":"https://github.com/ggml-org/llama.cpp/discussions/15396","nvidia":"https://www.nvidia.com/en-us/geforce/graphics-cards/compare/","hugging_face":"https://huggingface.co/models","reading_speed":"https://www.sciencedirect.com/science/article/abs/pii/S0749596X19300786"},"devices":[{"name":"M5 Max","kind":"apple","memory":[36,48,64,128],"vram":null,"bandwidth":614,"tg":119.92,"pp":3219.99,"measured":true,"old_build":false,"id":"m5-max","source":"https://github.com/ggml-org/llama.cpp/discussions/4167","variants":[{"bandwidth":460,"tg":93,"pp":2580,"measured":false,"memory":[36]}]},{"name":"M5 Ultra","kind":"apple","memory":[96,256,512],"vram":null,"bandwidth":1200,"tg":150,"pp":4900,"measured":false,"old_build":false,"id":"m5-ultra","source":"https://github.com/ggml-org/llama.cpp/discussions/4167","variants":[]},{"name":"M5 Pro","kind":"apple","memory":[24,48,64],"vram":null,"bandwidth":307,"tg":66.33,"pp":1340.44,"measured":true,"old_build":false,"id":"m5-pro","source":"https://github.com/ggml-org/llama.cpp/discussions/4167","variants":[]},{"name":"M6","kind":"apple","memory":[16,24,32],"vram":null,"bandwidth":170,"tg":34,"pp":940,"measured":false,"old_build":false,"id":"m6","source":"https://github.com/ggml-org/llama.cpp/discussions/4167","variants":[{"bandwidth":153,"tg":32,"memory":[16]}]},{"name":"M5","kind":"apple","memory":[16,24,32],"vram":null,"bandwidth":153,"tg":31.88,"pp":722.79,"measured":true,"old_build":false,"id":"m5","source":"https://github.com/ggml-org/llama.cpp/discussions/4167","variants":[]},{"name":"A18 Pro","kind":"apple","memory":[8],"vram":null,"bandwidth":60,"tg":12.5,"pp":85,"measured":false,"old_build":false,"id":"a18-pro","source":"https://github.com/ggml-org/llama.cpp/discussions/4167","variants":[]},{"name":"M3 Ultra","kind":"apple","memory":[96,256,512],"vram":null,"bandwidth":819,"tg":88.4,"pp":1073.09,"measured":true,"old_build":true,"id":"m3-ultra","source":"https://github.com/ggml-org/llama.cpp/discussions/4167","variants":[]},{"name":"M4 Max","kind":"apple","memory":[36,48,64,128],"vram":null,"bandwidth":546,"tg":83.06,"pp":885.68,"measured":true,"old_build":true,"id":"m4-max","source":"https://github.com/ggml-org/llama.cpp/discussions/4167","variants":[{"bandwidth":410,"tg":69.95,"pp":713.93,"memory":[36]}]},{"name":"M4 Pro","kind":"apple","memory":[24,48,64],"vram":null,"bandwidth":273,"tg":49.64,"pp":364.06,"measured":true,"old_build":true,"id":"m4-pro","source":"https://github.com/ggml-org/llama.cpp/discussions/4167","variants":[]},{"name":"M4","kind":"apple","memory":[16,24,32],"vram":null,"bandwidth":120,"tg":24.11,"pp":221.29,"measured":true,"old_build":true,"id":"m4","source":"https://github.com/ggml-org/llama.cpp/discussions/4167","variants":[]},{"name":"M3 Max","kind":"apple","memory":[36,48,64,96,128],"vram":null,"bandwidth":400,"tg":66.31,"pp":759.7,"measured":true,"old_build":true,"id":"m3-max","source":"https://github.com/ggml-org/llama.cpp/discussions/4167","variants":[{"bandwidth":300,"tg":56.58,"pp":567.59,"memory":[36,96]}]},{"name":"M3 Pro","kind":"apple","memory":[18,36],"vram":null,"bandwidth":150,"tg":30.65,"pp":269.49,"measured":true,"old_build":true,"id":"m3-pro","source":"https://github.com/ggml-org/llama.cpp/discussions/4167","variants":[]},{"name":"M3","kind":"apple","memory":[8,16,24],"vram":null,"bandwidth":100,"tg":21.34,"pp":186.75,"measured":true,"old_build":true,"id":"m3","source":"https://github.com/ggml-org/llama.cpp/discussions/4167","variants":[]},{"name":"M2 Ultra","kind":"apple","memory":[64,128,192],"vram":null,"bandwidth":800,"tg":88.64,"pp":1013.81,"measured":true,"old_build":true,"id":"m2-ultra","source":"https://github.com/ggml-org/llama.cpp/discussions/4167","variants":[]},{"name":"M2 Max","kind":"apple","memory":[32,64,96],"vram":null,"bandwidth":400,"tg":60.99,"pp":537.6,"measured":true,"old_build":true,"id":"m2-max","source":"https://github.com/ggml-org/llama.cpp/discussions/4167","variants":[]},{"name":"M2 Pro","kind":"apple","memory":[16,32],"vram":null,"bandwidth":200,"tg":37.87,"pp":294.24,"measured":true,"old_build":true,"id":"m2-pro","source":"https://github.com/ggml-org/llama.cpp/discussions/4167","variants":[]},{"name":"M2","kind":"apple","memory":[8,16,24],"vram":null,"bandwidth":100,"tg":21.7,"pp":145.91,"measured":true,"old_build":true,"id":"m2","source":"https://github.com/ggml-org/llama.cpp/discussions/4167","variants":[]},{"name":"M1 Ultra","kind":"apple","memory":[64,128],"vram":null,"bandwidth":800,"tg":74.93,"pp":772.24,"measured":true,"old_build":true,"id":"m1-ultra","source":"https://github.com/ggml-org/llama.cpp/discussions/4167","variants":[]},{"name":"M1 Max","kind":"apple","memory":[32,64],"vram":null,"bandwidth":400,"tg":54.61,"pp":400.26,"measured":true,"old_build":true,"id":"m1-max","source":"https://github.com/ggml-org/llama.cpp/discussions/4167","variants":[]},{"name":"M1 Pro","kind":"apple","memory":[16,32],"vram":null,"bandwidth":200,"tg":35.52,"pp":232.55,"measured":true,"old_build":true,"id":"m1-pro","source":"https://github.com/ggml-org/llama.cpp/discussions/4167","variants":[]},{"name":"M1","kind":"apple","memory":[8,16],"vram":null,"bandwidth":68.25,"tg":14.15,"pp":107.81,"measured":true,"old_build":true,"id":"m1","source":"https://github.com/ggml-org/llama.cpp/discussions/4167","variants":[]},{"name":"Intel MacBook Air, MacBook Pro 13″ (2018–2020)","kind":"intel_mac","memory":[8,16,32],"vram":null,"bandwidth":45,"tg":6.371517386519415,"pp":40.5,"measured":false,"old_build":false,"id":"intel-macbook","source":"https://github.com/ggml-org/llama.cpp/discussions/4167","variants":[]},{"name":"Intel MacBook Pro 15″, 16″ (2018–2019)","kind":"intel_mac","memory":[16,32,64],"vram":null,"bandwidth":40,"tg":5.673612543841551,"pp":36.0,"measured":false,"old_build":false,"id":"intel-macbook-pro-16","source":"https://github.com/ggml-org/llama.cpp/discussions/4167","variants":[]},{"name":"Intel iMac, Mac mini (2017–2020)","kind":"intel_mac","memory":[8,16,32,64],"vram":null,"bandwidth":42.7,"tg":6.050788232073558,"pp":38.43000000000001,"measured":false,"old_build":false,"id":"intel-imac","source":"https://github.com/ggml-org/llama.cpp/discussions/4167","variants":[]},{"name":"iMac Pro, Mac Pro (2017–2019)","kind":"intel_mac","memory":[32,64,128],"vram":null,"bandwidth":100,"tg":13.888538168228076,"pp":90.0,"measured":false,"old_build":false,"id":"intel-imac-pro","source":"https://github.com/ggml-org/llama.cpp/discussions/4167","variants":[]},{"name":"GeForce RTX 5090 (32 GB)","kind":"nvidia","memory":[16,32,64,128],"vram":32,"bandwidth":1792,"tg":300.4,"pp":14970.15,"measured":true,"old_build":false,"id":"rtx-5090","source":"https://github.com/ggml-org/llama.cpp/discussions/15013","variants":[]},{"name":"GeForce RTX 5080 (16 GB)","kind":"nvidia","memory":[16,32,64,128],"vram":16,"bandwidth":960,"tg":184.68,"pp":9487.7,"measured":true,"old_build":false,"id":"rtx-5080","source":"https://github.com/ggml-org/llama.cpp/discussions/15013","variants":[]},{"name":"GeForce RTX 5070 Ti (16 GB)","kind":"nvidia","memory":[16,32,64,128],"vram":16,"bandwidth":896,"tg":182.43,"pp":8419.56,"measured":true,"old_build":false,"id":"rtx-5070-ti","source":"https://github.com/ggml-org/llama.cpp/discussions/15013","variants":[]},{"name":"GeForce RTX 5070 (12 GB)","kind":"nvidia","memory":[16,32,64,128],"vram":12,"bandwidth":672,"tg":128.21,"pp":5783.44,"measured":true,"old_build":false,"id":"rtx-5070","source":"https://github.com/ggml-org/llama.cpp/discussions/15013","variants":[]},{"name":"GeForce RTX 5060 Ti (16 GB)","kind":"nvidia","memory":[16,32,64,128],"vram":16,"bandwidth":448,"tg":93.46,"pp":4195.53,"measured":true,"old_build":false,"id":"rtx-5060-ti-16","source":"https://github.com/ggml-org/llama.cpp/discussions/15013","variants":[]},{"name":"GeForce RTX 5060 Ti (8 GB)","kind":"nvidia","memory":[16,32,64,128],"vram":8,"bandwidth":448,"tg":93.46,"pp":4195.53,"measured":true,"old_build":false,"id":"rtx-5060-ti-8","source":"https://github.com/ggml-org/llama.cpp/discussions/15013","variants":[]},{"name":"GeForce RTX 5060 (8 GB)","kind":"nvidia","memory":[16,32,64,128],"vram":8,"bandwidth":448,"tg":93,"pp":3500,"measured":false,"old_build":false,"id":"rtx-5060","source":"https://github.com/ggml-org/llama.cpp/discussions/15013","variants":[]},{"name":"GeForce RTX 4090 (24 GB)","kind":"nvidia","memory":[16,32,64,128],"vram":24,"bandwidth":1008,"tg":188.96,"pp":14770.63,"measured":true,"old_build":false,"id":"rtx-4090","source":"https://github.com/ggml-org/llama.cpp/discussions/15013","variants":[]},{"name":"GeForce RTX 4080 SUPER (16 GB)","kind":"nvidia","memory":[16,32,64,128],"vram":16,"bandwidth":736,"tg":147.48,"pp":9439.01,"measured":true,"old_build":false,"id":"rtx-4080-super","source":"https://github.com/ggml-org/llama.cpp/discussions/15013","variants":[]},{"name":"GeForce RTX 4080 (16 GB)","kind":"nvidia","memory":[16,32,64,128],"vram":16,"bandwidth":716.8,"tg":143.47,"pp":9205.93,"measured":true,"old_build":false,"id":"rtx-4080","source":"https://github.com/ggml-org/llama.cpp/discussions/15013","variants":[]},{"name":"GeForce RTX 4070 Ti SUPER (16 GB)","kind":"nvidia","memory":[16,32,64,128],"vram":16,"bandwidth":672,"tg":132.85,"pp":7612.32,"measured":true,"old_build":false,"id":"rtx-4070-ti-super","source":"https://github.com/ggml-org/llama.cpp/discussions/15013","variants":[]},{"name":"GeForce RTX 4070 Ti (12 GB)","kind":"nvidia","memory":[16,32,64,128],"vram":12,"bandwidth":504,"tg":110.53,"pp":4981.44,"measured":true,"old_build":false,"id":"rtx-4070-ti","source":"https://github.com/ggml-org/llama.cpp/discussions/10879","variants":[]},{"name":"GeForce RTX 4070 SUPER (12 GB)","kind":"nvidia","memory":[16,32,64,128],"vram":12,"bandwidth":504,"tg":108.74,"pp":4608.2,"measured":true,"old_build":false,"id":"rtx-4070-super","source":"https://github.com/ggml-org/llama.cpp/discussions/10879","variants":[]},{"name":"GeForce RTX 4070 (12 GB)","kind":"nvidia","memory":[16,32,64,128],"vram":12,"bandwidth":504,"tg":92.29,"pp":3179.37,"measured":true,"old_build":false,"id":"rtx-4070","source":"https://github.com/ggml-org/llama.cpp/discussions/10879","variants":[]},{"name":"GeForce RTX 4060 Ti (16 GB)","kind":"nvidia","memory":[16,32,64,128],"vram":16,"bandwidth":288,"tg":64.03,"pp":3803.45,"measured":true,"old_build":false,"id":"rtx-4060-ti-16","source":"https://github.com/ggml-org/llama.cpp/discussions/15013","variants":[]},{"name":"GeForce RTX 4060 Ti (8 GB)","kind":"nvidia","memory":[16,32,64,128],"vram":8,"bandwidth":288,"tg":64.03,"pp":3803.45,"measured":true,"old_build":false,"id":"rtx-4060-ti-8","source":"https://github.com/ggml-org/llama.cpp/discussions/15013","variants":[]},{"name":"GeForce RTX 4060 (8 GB)","kind":"nvidia","memory":[16,32,64,128],"vram":8,"bandwidth":272,"tg":59,"pp":2700,"measured":false,"old_build":false,"id":"rtx-4060","source":"https://github.com/ggml-org/llama.cpp/discussions/15013","variants":[]},{"name":"GeForce RTX 3090 (24 GB)","kind":"nvidia","memory":[16,32,64,128],"vram":24,"bandwidth":936,"tg":161.89,"pp":5560.06,"measured":true,"old_build":false,"id":"rtx-3090","source":"https://github.com/ggml-org/llama.cpp/discussions/15013","variants":[]},{"name":"GeForce RTX 3080 (10 GB)","kind":"nvidia","memory":[16,32,64,128],"vram":10,"bandwidth":760,"tg":139.95,"pp":5569.56,"measured":true,"old_build":false,"id":"rtx-3080","source":"https://github.com/ggml-org/llama.cpp/discussions/15013","variants":[]},{"name":"GeForce RTX 3070 (8 GB)","kind":"nvidia","memory":[16,32,64,128],"vram":8,"bandwidth":448,"tg":78.71,"pp":2113.02,"measured":true,"old_build":false,"id":"rtx-3070","source":"https://github.com/ggml-org/llama.cpp/discussions/10879","variants":[]},{"name":"GeForce RTX 3060 Ti (8 GB)","kind":"nvidia","memory":[16,32,64,128],"vram":8,"bandwidth":448,"tg":86.74,"pp":2361.43,"measured":true,"old_build":false,"id":"rtx-3060-ti","source":"https://github.com/ggml-org/llama.cpp/discussions/10879","variants":[]},{"name":"GeForce RTX 3060 (12 GB)","kind":"nvidia","memory":[16,32,64,128],"vram":12,"bandwidth":360,"tg":76.92,"pp":2407.67,"measured":true,"old_build":false,"id":"rtx-3060","source":"https://github.com/ggml-org/llama.cpp/discussions/15013","variants":[]},{"name":"Radeon RX 9070 XT (16 GB)","kind":"amd","memory":[16,32,64,128],"vram":16,"bandwidth":640,"tg":137.11,"pp":5036.04,"measured":true,"old_build":false,"id":"rx-9070-xt","source":"https://github.com/ggml-org/llama.cpp/discussions/10879","variants":[]},{"name":"Radeon RX 9070 (16 GB)","kind":"amd","memory":[16,32,64,128],"vram":16,"bandwidth":640,"tg":119.71,"pp":3164.1,"measured":true,"old_build":false,"id":"rx-9070","source":"https://github.com/ggml-org/llama.cpp/discussions/10879","variants":[]},{"name":"Radeon RX 9060 XT (16 GB)","kind":"amd","memory":[16,32,64,128],"vram":16,"bandwidth":320,"tg":70.54,"pp":2141.67,"measured":true,"old_build":false,"id":"rx-9060-xt-16","source":"https://github.com/ggml-org/llama.cpp/discussions/10879","variants":[]},{"name":"Radeon RX 9060 XT (8 GB)","kind":"amd","memory":[16,32,64,128],"vram":8,"bandwidth":320,"tg":70.54,"pp":2141.67,"measured":true,"old_build":false,"id":"rx-9060-xt-8","source":"https://github.com/ggml-org/llama.cpp/discussions/10879","variants":[]},{"name":"Radeon RX 7900 XTX (24 GB)","kind":"amd","memory":[16,32,64,128],"vram":24,"bandwidth":960,"tg":182.63,"pp":3726.99,"measured":true,"old_build":false,"id":"rx-7900-xtx","source":"https://github.com/ggml-org/llama.cpp/discussions/10879","variants":[]},{"name":"Radeon RX 7900 XT (20 GB)","kind":"amd","memory":[16,32,64,128],"vram":20,"bandwidth":800,"tg":123.18,"pp":2941.58,"measured":true,"old_build":false,"id":"rx-7900-xt","source":"https://github.com/ggml-org/llama.cpp/discussions/10879","variants":[]},{"name":"Radeon RX 7900 GRE (16 GB)","kind":"amd","memory":[16,32,64,128],"vram":16,"bandwidth":576,"tg":116.11,"pp":2336.31,"measured":true,"old_build":false,"id":"rx-7900-gre","source":"https://github.com/ggml-org/llama.cpp/discussions/10879","variants":[]},{"name":"Radeon RX 7800 XT (16 GB)","kind":"amd","memory":[16,32,64,128],"vram":16,"bandwidth":624,"tg":118.27,"pp":2017.33,"measured":true,"old_build":false,"id":"rx-7800-xt","source":"https://github.com/ggml-org/llama.cpp/discussions/10879","variants":[]},{"name":"Radeon RX 7600 XT (16 GB)","kind":"amd","memory":[16,32,64,128],"vram":16,"bandwidth":288,"tg":53.02,"pp":840.85,"measured":true,"old_build":false,"id":"rx-7600-xt","source":"https://github.com/ggml-org/llama.cpp/discussions/10879","variants":[]},{"name":"Radeon RX 7600 (8 GB)","kind":"amd","memory":[16,32,64,128],"vram":8,"bandwidth":288,"tg":58.03,"pp":1167.2,"measured":true,"old_build":false,"id":"rx-7600","source":"https://github.com/ggml-org/llama.cpp/discussions/10879","variants":[]},{"name":"AMD Ryzen AI Max+ 395","kind":"shared","memory":[64,128],"vram":0.75,"bandwidth":256,"tg":53.59,"pp":1288.96,"measured":true,"old_build":false,"id":"ryzen-ai-max","source":"https://github.com/ggml-org/llama.cpp/discussions/10879","variants":[]},{"name":"NVIDIA DGX Spark","kind":"shared","memory":[128],"vram":0.9,"bandwidth":273,"tg":56.74,"pp":3661.37,"measured":true,"old_build":false,"id":"dgx-spark","source":"https://github.com/ggml-org/llama.cpp/discussions/15013","variants":[]},{"name":"Intel Core Ultra 300 (integrated graphics)","kind":"shared","memory":[16,32,64],"vram":0.5,"bandwidth":120,"tg":24.45,"pp":1380.07,"measured":true,"old_build":false,"id":"core-ultra-300","source":"https://github.com/ggml-org/llama.cpp/discussions/10879","variants":[]},{"name":"Intel Core Ultra 200 (integrated graphics)","kind":"shared","memory":[16,32,64],"vram":0.5,"bandwidth":100,"tg":24.37,"pp":864.99,"measured":true,"old_build":false,"id":"core-ultra-200","source":"https://github.com/ggml-org/llama.cpp/discussions/10879","variants":[]},{"name":"AMD Ryzen AI 300 (integrated graphics)","kind":"shared","memory":[16,32,64],"vram":0.5,"bandwidth":100,"tg":22.41,"pp":479.07,"measured":true,"old_build":false,"id":"ryzen-ai-300","source":"https://github.com/ggml-org/llama.cpp/discussions/10879","variants":[]},{"name":"PC without a graphics card (DDR5 memory)","kind":"cpu","memory":[8,16,32,64,128],"vram":null,"bandwidth":89.6,"tg":12.489229053677327,"pp":80.64,"measured":false,"old_build":false,"id":"cpu-ddr5","source":"https://github.com/ggml-org/llama.cpp/discussions/4167","variants":[]},{"name":"PC without a graphics card (DDR4 memory)","kind":"cpu","memory":[8,16,32,64,128],"vram":null,"bandwidth":51.2,"tg":7.233496018494735,"pp":46.080000000000005,"measured":false,"old_build":false,"id":"cpu-ddr4","source":"https://github.com/ggml-org/llama.cpp/discussions/4167","variants":[]}],"models":[{"name":"Qwen3 4B","publisher":"Qwen (Alibaba)","released":"2025-04","params":4.0,"active":null,"layers":36,"kv_kib":144,"context":32,"license":"Apache 2.0","q4":2.497,"q8":4.28,"arch":"dense","tier":"everyday","one_click":true,"id":"qwen3-4b","url":"https://huggingface.co/Qwen/Qwen3-4B"},{"name":"Qwen3 8B","publisher":"Qwen (Alibaba)","released":"2025-04","params":8.2,"active":null,"layers":36,"kv_kib":144,"context":32,"license":"Apache 2.0","q4":5.028,"q8":8.71,"arch":"dense","tier":"solid","one_click":true,"id":"qwen3-8b","url":"https://huggingface.co/Qwen/Qwen3-8B"},{"name":"Qwen3 14B","publisher":"Qwen (Alibaba)","released":"2025-04","params":14.8,"active":null,"layers":40,"kv_kib":160,"context":32,"license":"Apache 2.0","q4":9.002,"q8":15.7,"arch":"dense","tier":"solid","one_click":true,"id":"qwen3-14b","url":"https://huggingface.co/Qwen/Qwen3-14B"},{"name":"Qwen3 0.6B","publisher":"Qwen (Alibaba)","released":"2025-04","params":0.6,"active":null,"layers":28,"kv_kib":112,"context":32,"license":"Apache 2.0","q4":0.4,"q8":0.64,"arch":"dense","tier":"basic","one_click":false,"id":"qwen3-0-6b","url":"https://huggingface.co/Qwen/Qwen3-0.6B"},{"name":"Qwen3 1.7B","publisher":"Qwen (Alibaba)","released":"2025-04","params":1.7,"active":null,"layers":28,"kv_kib":112,"context":32,"license":"Apache 2.0","q4":1.11,"q8":1.83,"arch":"dense","tier":"basic","one_click":false,"id":"qwen3-1-7b","url":"https://huggingface.co/Qwen/Qwen3-1.7B"},{"name":"Qwen3 32B","publisher":"Qwen (Alibaba)","released":"2025-04","params":32.8,"active":null,"layers":64,"kv_kib":256,"context":32,"license":"Apache 2.0","q4":19.76,"q8":34.82,"arch":"dense","tier":"strong","one_click":false,"id":"qwen3-32b","url":"https://huggingface.co/Qwen/Qwen3-32B"},{"name":"Qwen3 30B-A3B","publisher":"Qwen (Alibaba)","released":"2025-04","params":30.5,"active":3.3,"layers":48,"kv_kib":96,"context":32,"license":"Apache 2.0","q4":18.56,"q8":32.48,"arch":"moe","tier":"strong","one_click":false,"id":"qwen3-30b-a3b","url":"https://huggingface.co/Qwen/Qwen3-30B-A3B"},{"name":"Qwen3-Coder 30B-A3B","publisher":"Qwen (Alibaba)","released":"2025-07","params":30.5,"active":3.3,"layers":48,"kv_kib":96,"context":256,"license":"Apache 2.0","q4":18.56,"q8":32.48,"arch":"moe","tier":"strong","one_click":false,"id":"qwen3-coder-30b-a3b","url":"https://huggingface.co/Qwen/Qwen3-Coder-30B-A3B-Instruct"},{"name":"Qwen3.5 2B","publisher":"Qwen (Alibaba)","released":"2026-02","params":2.0,"active":null,"layers":24,"kv_kib":12,"context":262,"license":"Apache 2.0","q4":1.28,"q8":2.01,"arch":"hybrid","tier":"basic","one_click":false,"id":"qwen3-5-2b","url":"https://huggingface.co/Qwen/Qwen3.5-2B"},{"name":"Qwen3.5 4B","publisher":"Qwen (Alibaba)","released":"2026-02","params":4.66,"active":null,"layers":32,"kv_kib":32,"context":262,"license":"Apache 2.0","q4":2.74,"q8":4.48,"arch":"hybrid","tier":"everyday","one_click":false,"id":"qwen3-5-4b","url":"https://huggingface.co/Qwen/Qwen3.5-4B"},{"name":"Qwen3.5 9B","publisher":"Qwen (Alibaba)","released":"2026-02","params":9.65,"active":null,"layers":32,"kv_kib":32,"context":262,"license":"Apache 2.0","q4":5.68,"q8":9.53,"arch":"hybrid","tier":"solid","one_click":false,"id":"qwen3-5-9b","url":"https://huggingface.co/Qwen/Qwen3.5-9B"},{"name":"Qwen3.8 27B","publisher":"Qwen (Alibaba)","released":"2026-08","params":27.8,"active":null,"layers":64,"kv_kib":64,"context":262,"license":"Apache 2.0","q4":16.46,"q8":29.05,"arch":"hybrid","tier":"strong","one_click":false,"id":"qwen3-8-27b","url":"https://huggingface.co/Qwen/Qwen3.8-27B"},{"name":"Qwen3.6 35B-A3B","publisher":"Qwen (Alibaba)","released":"2026-04","params":36.0,"active":3.0,"layers":40,"kv_kib":20,"context":262,"license":"Apache 2.0","q4":22.13,"q8":36.9,"arch":"hybrid_moe","tier":"strong","one_click":false,"id":"qwen3-6-35b-a3b","url":"https://huggingface.co/Qwen/Qwen3.6-35B-A3B"},{"name":"Qwen3-Next 80B-A3B","publisher":"Qwen (Alibaba)","released":"2025-09","params":80.0,"active":3.0,"layers":48,"kv_kib":24,"context":262,"license":"Apache 2.0","q4":48.5,"q8":84.8,"arch":"hybrid_moe","tier":"very_strong","one_click":false,"id":"qwen3-next-80b-a3b","url":"https://huggingface.co/Qwen/Qwen3-Next-80B-A3B-Instruct"},{"name":"Qwen3.5 122B-A10B","publisher":"Qwen (Alibaba)","released":"2026-02","params":122.0,"active":10.0,"layers":48,"kv_kib":24,"context":262,"license":"Apache 2.0","q4":76.54,"q8":129.87,"arch":"hybrid_moe","tier":"very_strong","one_click":false,"id":"qwen3-5-122b-a10b","url":"https://huggingface.co/Qwen/Qwen3.5-122B-A10B"},{"name":"Qwen3 235B-A22B","publisher":"Qwen (Alibaba)","released":"2025-07","params":235.0,"active":22.0,"layers":94,"kv_kib":188,"context":256,"license":"Apache 2.0","q4":142.15,"q8":249.94,"arch":"moe","tier":"frontier","one_click":false,"id":"qwen3-235b-a22b","url":"https://huggingface.co/Qwen/Qwen3-235B-A22B-Instruct-2507"},{"name":"gpt-oss-20b","publisher":"OpenAI","released":"2025-08","params":20.9,"active":3.6,"layers":24,"kv_kib":24,"context":128,"license":"Apache 2.0","q4":12.11,"q8":null,"arch":"moe","tier":"strong","one_click":false,"id":"gpt-oss-20b","url":"https://huggingface.co/openai/gpt-oss-20b"},{"name":"gpt-oss-120b","publisher":"OpenAI","released":"2025-08","params":116.8,"active":5.1,"layers":36,"kv_kib":36,"context":128,"license":"Apache 2.0","q4":63.39,"q8":null,"arch":"moe","tier":"very_strong","one_click":false,"id":"gpt-oss-120b","url":"https://huggingface.co/openai/gpt-oss-120b"},{"name":"Gemma 4 E2B","publisher":"Google","released":"2026-03","params":5.1,"active":2.3,"layers":35,"kv_kib":6,"context":128,"license":"Apache 2.0","q4":3.11,"q8":5.15,"arch":"dense","tier":"basic","one_click":false,"id":"gemma4-e2b","url":"https://huggingface.co/google/gemma-4-E2B-it"},{"name":"Gemma 4 E4B","publisher":"Google","released":"2026-03","params":8.0,"active":4.5,"layers":42,"kv_kib":16,"context":128,"license":"Apache 2.0","q4":4.98,"q8":8.29,"arch":"dense","tier":"everyday","one_click":false,"id":"gemma4-e4b","url":"https://huggingface.co/google/gemma-4-E4B-it"},{"name":"Gemma 4 12B","publisher":"Google","released":"2026-05","params":12.0,"active":null,"layers":48,"kv_kib":8,"context":256,"license":"Apache 2.0","q4":7.12,"q8":13.13,"arch":"dense","tier":"solid","one_click":false,"id":"gemma4-12b","url":"https://huggingface.co/google/gemma-4-12B-it"},{"name":"Gemma 4 26B-A4B","publisher":"Google","released":"2026-03","params":25.8,"active":4.0,"layers":30,"kv_kib":10,"context":256,"license":"Apache 2.0","q4":16.95,"q8":26.86,"arch":"moe","tier":"strong","one_click":false,"id":"gemma4-26b-a4b","url":"https://huggingface.co/google/gemma-4-26B-A4B-it"},{"name":"Gemma 4 31B","publisher":"Google","released":"2026-03","params":31.3,"active":null,"layers":60,"kv_kib":40,"context":256,"license":"Apache 2.0","q4":18.32,"q8":33.15,"arch":"dense","tier":"strong","one_click":false,"id":"gemma4-31b","url":"https://huggingface.co/google/gemma-4-31B-it"},{"name":"Llama 3.2 3B","publisher":"Meta","released":"2024-09","params":3.2,"active":null,"layers":28,"kv_kib":112,"context":128,"license":"Llama 3.2 Community","q4":2.02,"q8":3.42,"arch":"dense","tier":"everyday","one_click":false,"id":"llama3-2-3b","url":"https://huggingface.co/meta-llama/Llama-3.2-3B-Instruct"},{"name":"Llama 3.1 8B","publisher":"Meta","released":"2024-07","params":8.0,"active":null,"layers":32,"kv_kib":128,"context":128,"license":"Llama 3.1 Community","q4":4.92,"q8":8.54,"arch":"dense","tier":"solid","one_click":false,"id":"llama3-1-8b","url":"https://huggingface.co/meta-llama/Llama-3.1-8B-Instruct"},{"name":"Llama 3.3 70B","publisher":"Meta","released":"2024-12","params":70.6,"active":null,"layers":80,"kv_kib":320,"context":128,"license":"Llama 3.3 Community","q4":42.52,"q8":74.98,"arch":"dense","tier":"very_strong","one_click":false,"id":"llama3-3-70b","url":"https://huggingface.co/meta-llama/Llama-3.3-70B-Instruct"},{"name":"Ministral 3 8B","publisher":"Mistral AI","released":"2025-12","params":8.9,"active":null,"layers":34,"kv_kib":136,"context":256,"license":"Apache 2.0","q4":5.2,"q8":9.03,"arch":"dense","tier":"solid","one_click":false,"id":"ministral3-8b","url":"https://huggingface.co/mistralai/Ministral-3-8B-Instruct-2512"},{"name":"Ministral 3 14B","publisher":"Mistral AI","released":"2025-12","params":13.9,"active":null,"layers":40,"kv_kib":160,"context":256,"license":"Apache 2.0","q4":8.24,"q8":14.36,"arch":"dense","tier":"solid","one_click":false,"id":"ministral3-14b","url":"https://huggingface.co/mistralai/Ministral-3-14B-Instruct-2512"},{"name":"Mistral Small 3.2 24B","publisher":"Mistral AI","released":"2025-06","params":24.0,"active":null,"layers":40,"kv_kib":160,"context":128,"license":"Apache 2.0","q4":14.33,"q8":25.05,"arch":"dense","tier":"strong","one_click":false,"id":"mistral-small-3-2","url":"https://huggingface.co/mistralai/Mistral-Small-3.2-24B-Instruct-2506"},{"name":"Devstral Small 2 24B","publisher":"Mistral AI","released":"2025-12","params":24.0,"active":null,"layers":40,"kv_kib":160,"context":384,"license":"Apache 2.0","q4":14.33,"q8":25.05,"arch":"dense","tier":"strong","one_click":false,"id":"devstral-small-2","url":"https://huggingface.co/mistralai/Devstral-Small-2-24B-Instruct-2512"},{"name":"Mistral Small 4 119B","publisher":"Mistral AI","released":"2026-03","params":119.0,"active":6.5,"layers":36,"kv_kib":22.5,"context":256,"license":"Apache 2.0","q4":71.81,"q8":126.47,"arch":"moe","tier":"very_strong","one_click":false,"id":"mistral-small-4","url":"https://huggingface.co/mistralai/Mistral-Small-4-119B-2603"},{"name":"Phi-4-mini","publisher":"Microsoft","released":"2025-02","params":3.8,"active":null,"layers":32,"kv_kib":128,"context":128,"license":"MIT","q4":2.49,"q8":4.08,"arch":"dense","tier":"everyday","one_click":false,"id":"phi4-mini","url":"https://huggingface.co/microsoft/Phi-4-mini-instruct"},{"name":"Phi-4 14B","publisher":"Microsoft","released":"2024-12","params":14.7,"active":null,"layers":40,"kv_kib":200,"context":16,"license":"MIT","q4":8.89,"q8":15.58,"arch":"dense","tier":"solid","one_click":false,"id":"phi4","url":"https://huggingface.co/microsoft/phi-4"},{"name":"DeepSeek-R1-0528-Qwen3-8B","publisher":"DeepSeek","released":"2025-05","params":8.2,"active":null,"layers":36,"kv_kib":144,"context":128,"license":"MIT","q4":5.03,"q8":8.71,"arch":"dense","tier":"solid","one_click":false,"id":"deepseek-r1-qwen3-8b","url":"https://huggingface.co/deepseek-ai/DeepSeek-R1-0528-Qwen3-8B"},{"name":"GLM-4.7-Flash","publisher":"Z.ai","released":"2026-01","params":31.2,"active":3.0,"layers":47,"kv_kib":53,"context":200,"license":"MIT","q4":18.31,"q8":31.84,"arch":"moe","tier":"strong","one_click":false,"id":"glm-4-7-flash","url":"https://huggingface.co/zai-org/GLM-4.7-Flash"},{"name":"MiniCPM5 2B","publisher":"OpenBMB","released":"2026-09","params":2.5,"active":null,"layers":42,"kv_kib":42,"context":128,"license":"Apache 2.0","q4":1.56,"q8":2.68,"arch":"dense","tier":"basic","one_click":false,"id":"minicpm5-2b","url":"https://huggingface.co/openbmb/MiniCPM5-2B"}]}