VRAMGlass contains affiliate links. As an eBay Partner, we may be compensated if you make a purchase.As an eBay Partner, we may be compensated.Learn more

GPUs tracked57Price history49 daysPrice points (24h)39,704 pointsBiggest 24h dropRadeon RX 7900 XTX-3.8%Biggest 24h riseCMP 170HX 40GB+19.8%Updated
$ USD

Popular Open-Weight Models for Local Deployment

45 open-weight model versions across 9 families tracked as of Oct 21, 2026, with parameter counts, quantized weight sizes and the VRAM needed to run them.

VersionVendorLicense
DeepSeek V4.1 FlashDeepSeekOct 17, 20267638 / 16508444.7264.5—est.MIT
Muse Spark OpenMeta AIOct 15, 202681642871.7505.5311.2208.9Muse Model License
Hy4 previewTencentAug 28, 202677049857est.467.3307.2est.235.4Apache-2.0
GLM 5.3Z.ai / Zhipu AIAug 28, 202674440801.4467.3253.9228.5GLM-5.3 License
GLM 5.3 FlashZ.ai / Zhipu AIAug 26, 202632018341199.7108.797.6MIT
Qwen3.8 Flash NextAlibabaAug 26, 20261766188.2111.378.974.5Qwen Community License 1.0
Qwen3.8 27BAlibabaAug 14, 202627.827.82916.59.86.7Apache-2.0
DeepSeek V4 ProDeepSeekAug 13, 20261,60049873.4849.7638.4est.403.2est.MIT
Qwen3.8 2.4T A95BAlibabaAug 12, 20262,446952,600.21,310.9730.7564Qwen3.8-Max License
DeepSeek V4 FlashDeepSeekJul 31, 202628413161.9155.196.886.9MIT
Kimi K3Moonshot AIJul 27, 20262,8001041,561.21,508.7861.3648.9Kimi K3 License
Hy3TencentJul 6, 202629521317.7182.2107.271.1Apache-2.0
GLM 5.2Z.ai / Zhipu AIJun 16, 202674440801.4465.8253.9228.5MIT
Kimi K2.7 CodeMoonshot AIJun 12, 20261,00032594.5583.7339.5303.9Modified MIT
Gemma 4 12BGoogle DeepMindJun 3, 2026121212.77.14.73est.Apache-2.0
Qwen3.6 27BAlibabaApr 22, 202627.827.828.616.811.87est.Apache-2.0
Kimi K2.6Moonshot AIApr 20, 20261,00032594.5583.7340.5252est.Modified MIT
Qwen3.6 35B A3BAlibabaApr 16, 202636336.922.112.310Apache-2.0
Gemma 4 31BGoogle DeepMindApr 2, 202630.730.732.618.311.87.7est.Apache-2.0
Gemma 4 26B A4BGoogle DeepMindApr 2, 202625.23.826.916.910.56.4est.Apache-2.0
Gemma 4 E4BGoogle DeepMindApr 2, 2026888.253.82est.Apache-2.0
Qwen3.5 9BAlibabaMar 2, 20269.79.79.55.74.12.4est.Apache-2.0
Kimi K2.5Moonshot AIJan 27, 20261,000321,091.1621.2374.8300.6Modified MIT
GLM 4.7 FlashZ.ai / Zhipu AIJan 19, 202630331.818.311.99.8MIT
Ministral 3 14BMistral AIDec 2, 202513.513.514.48.25.53.7Apache-2.0
Ministral 3 8BMistral AIDec 2, 20258.48.495.23.62.4Apache-2.0
DeepSeek V3.2DeepSeekDec 1, 202567137713.3405.4246.6199.3MIT
Kimi Linear 48B A3BMoonshot AIOct 30, 202549.1352.230.117.910.9MIT
gpt-oss Safeguard 20BOpenAIOct 29, 202520.93.612.111.68.3est.5.3est.Apache-2.0
Kimi K2Moonshot AISep 3, 20251,000321,091.1621.3383.6304.8Modified MIT
gpt-oss 120BOpenAIAug 5, 2025116.85.163.462.846.6est.29.4est.Apache-2.0
gpt-oss 20BOpenAIAug 5, 202520.93.612.111.68.3est.5.3est.Apache-2.0
GLM 4.5 AirZ.ai / Zhipu AIJul 28, 202510612117.57347.440.1MIT
Mistral Small 3.2 24BMistral AIJun 19, 2025242425.114.39.56.1est.Apache-2.0
DeepSeek R1 Qwen3 8BDeepSeekMay 29, 20258.28.28.753.52.4MIT
Qwen3 32BAlibabaApr 27, 202532.832.834.819.812.88.3Apache-2.0
Qwen3 8BAlibabaApr 27, 20258.28.28.753.52.4Apache-2.0
Llama 4 Scout 17B 16EMeta AIApr 5, 2025108.617114.565.442.435Llama 4 Community License
DeepSeek R1 Distill Qwen 32BDeepSeekJan 20, 202532.832.834.819.913.18.3est.MIT
DeepSeek R1 Distill Qwen 14BDeepSeekJan 20, 202514.814.815.796.53.7est.MIT
Llama 3.3 70BMeta AIDec 6, 202470.670.67542.527.416.8Llama 3.3 Community License
Llama 3.2 3BMeta AISep 25, 20243.23.23.421.3est.0.8est.Llama 3.2 Community License
Llama 3.1 8BMeta AIJul 23, 2024888.54.93.72est.Llama 3.1 Community License
Mistral NemoMistral AIJul 17, 202412.312.3137.55.43.1est.Apache-2.0
Mistral 7B v0.3Mistral AIMay 22, 20247.37.37.74.42.51.8Apache-2.0

Each size is the GGUF file of that bit width in the repository (Q8_0 for 8-bit, unsloth dynamic quants preferred for the rest); missing tiers are estimated from bytes per parameter and tagged · Prices are median asking prices, not sold prices · Methodology