RunsOnMy ModelsGPUs & MacsBy VRAMBest GPUCan I run…?API pricesVRAM calculatorNew

New this week: can you run them?

Open models released on Hugging Face in the last 7 days that are getting real attention (likes and trending), checked against 65 GPUs and Macs. The list refreshes every hour, so new releases usually show up here within an hour or two of picking up traction. Last checked 2026-11-17 23:16 UTC.

ModelReleasedParams4-bit · 8KVRAMCheapest GPUSmallest MacLikes
SenseNova Loop 9B sensenovaNov 17, 20269.4B6.4 GBruns on 8 GB+ VRAMArc B580Mac M4 16GB175
KORMo Web 11B KORMo-VLNov 14, 202611B8.3 GBruns on 12 GB+ VRAMArc B580Mac M4 16GB73
SingVL S1 4B inclusionAINov 11, 20264.5B3.6 GBruns on 8 GB+ VRAMArc B580Mac M4 16GB5

Earlier this month

ModelReleasedParams4-bit · 8KCheapest GPU
Nex N2.6 nex-agiNov 10, 2026321B188 GBnone in VRAM
MiniMax H3 Image PE MiniMaxAINov 8, 202627B16.8 GBRX 7900 XT
Arete 27B NousResearchNov 6, 202627B17.4 GBRX 7900 XTX
Inkling 2 Small thinkingmachinesNov 4, 2026594B346 GBnone in VRAM
LongCat 2.5 meituan-longcatNov 4, 20261.78T1014 GBnone in VRAM
Kimi K3.1 Pro moonshotaiNov 2, 20262.78T1584 GBnone in VRAM
Mage VL Cowork microsoftOct 30, 20264.7B4.3 GBArc B580
Raon 1 Nano KRAFTONOct 29, 20263.5B3.0 GBArc B580
SenseNova Loop 9B Preview sensenovaOct 28, 20269.7B6.4 GBArc B580
Ling 3.1 Flash 1019 inclusionAIOct 27, 2026572B335 GBnone in VRAM
LFM3 1.7B Base LiquidAIOct 26, 20261.7B1.7 GBArc B580
Unlimited OCR 2 baiduOct 26, 20263.3B3.0 GBArc B580
d1 2B LiquidAIOct 24, 20261.9B1.8 GBArc B580
LFM3 1.7B LiquidAIOct 23, 20261.7B1.8 GBArc B580
WeVisOne VL 4B tencentOct 22, 20264.8B4.3 GBArc B580

Everything new this month, plus new API models → · RSS

How a model gets listed

Every hour we check Hugging Face's trending lists and the newest uploads from the AI labs we track. A release from the last 14 days is listed when it's from a tracked lab or has at least 20 likes or a trending score of 15. Finetunes from new accounts need 60 likes. Quantized re-uploads, merges and "uncensored" variants are skipped. Each listed model gets a model page, a "Can I run it?" page, GPU-by-GPU pages, a sitemap entry and an IndexNow ping. Numbers come from the model's own config and real GGUF sizes once they're published. Methodology.