我的 GPU 能运行这个模型吗？

本页面由 AI 自动撰写和翻译，并由软件对照来源进行检查，没有人工编辑。规格来自所列来源；显存需求和速度是按已公布的公式计算出的估算值。 我们如何使用 AI

一个数据库，记录哪些开放权重语言模型能装进哪些 GPU 的显存——使用哪种 GGUF 量化、多大的上下文，以及最快能达到多少速度。只有在两个独立的公开来源一致时才显示规格，每个估算都会给出其公式。

数据库中共有 389 个模型和 60 款显卡。 数值更新于 2026年10月19日

快速查询

选择一款 GPU 和一个模型即可查看结果。计算器还支持多张相同的显卡、其他上下文长度和 KV 缓存类型。

GPUAMD Instinct MI210AMD Instinct MI300XAMD Instinct MI325XAMD Instinct MI350XAMD Instinct MI355XAMD Radeon AI PRO R9700AMD Radeon Pro V520AMD Radeon PRO W7800 32GBAMD Radeon PRO W7800 48GBAMD Radeon PRO W7900AMD Radeon RX 7700 XTAMD Radeon RX 7800 XTAMD Radeon RX 7900 GREAMD Radeon RX 7900 XTAMD Radeon RX 7900 XTXNVIDIA B100NVIDIA B200NVIDIA GeForce GTX 1050 TiNVIDIA GeForce GTX 1060NVIDIA GeForce GTX 1070NVIDIA GeForce GTX 1070 TiNVIDIA GeForce GTX 1080NVIDIA GeForce GTX 1080 TiNVIDIA GeForce GTX 1650NVIDIA GeForce GTX 1660NVIDIA GeForce GTX 1660 SUPERNVIDIA GeForce RTX 2060 6GBNVIDIA GeForce RTX 2060 SUPERNVIDIA GeForce RTX 2070NVIDIA GeForce RTX 2080NVIDIA GeForce RTX 2080 TiNVIDIA GeForce RTX 3050 8GBNVIDIA GeForce RTX 3060 TiNVIDIA GeForce RTX 3070NVIDIA GeForce RTX 3070 TiNVIDIA GeForce RTX 3080 10GBNVIDIA GeForce RTX 3080 TiNVIDIA GeForce RTX 3090NVIDIA GeForce RTX 3090 TiNVIDIA GeForce RTX 4060NVIDIA GeForce RTX 4060 Ti 16GBNVIDIA GeForce RTX 4070NVIDIA GeForce RTX 4070 TiNVIDIA GeForce RTX 4070 Ti SUPERNVIDIA GeForce RTX 4080NVIDIA GeForce RTX 5060NVIDIA GeForce RTX 5060 Ti 16GBNVIDIA GeForce RTX 5070NVIDIA GeForce RTX 5070 TiNVIDIA GeForce RTX 5080NVIDIA GeForce RTX 5090NVIDIA H100 NVLNVIDIA H100 PCIeNVIDIA H200 SXMNVIDIA L4NVIDIA L40NVIDIA P100NVIDIA P4NVIDIA Tesla P40NVIDIA Tesla V100 32GB

模型Qwen3-0.6BQwen3-4Bgemma-4-26B-A4B-itQwen3-8Bgemma-4-31B-itQwen3-14BQwen3.5-4BQwen3.5-9BQwen3.8-27BQwen2.5-7B-InstructGLM-5.3-FlashQwen3-VL-8B-Instructgpt-oss-20bLlama-3.1-8B-InstructDeepSeek-V3.2Llama-3.2-1B-InstructQwen2.5-VL-7B-InstructQwen2.5-3B-Instructgemma-4-E4B-itQwen3-1.7BQwen3.6-35B-A3BQwen3-VL-4B-Instructgpt-oss-120bQwen3.5-0.8BQwen3-32BQwen3-4B-Instruct-2507Qwen3.5-2Bgemma-4-E2B-itQwen3.6-27BQwen3-VL-2B-InstructGLM-5.3Qwen3.5-27Bgemma-3-1b-itMistral-7B-Instruct-v0.2gemma-4-12B-itNVIDIA-Nemotron-3-Nano-4B-BF16Qwen2.5-Coder-7B-InstructQwen2.5-32B-InstructQwen3.5-35B-A3BDeepSeek-V3MiniCPM5-2BDeepSeek-R1Llama-3.2-3B-InstructNVIDIA-Nemotron-3-Nano-30B-A3B-BF16GLM-4.7-FlashQwen3-30B-A3Bgemma-3-4b-itQwen3-30B-A3B-Instruct-2507NVIDIA-Nemotron-3-Super-120B-A12B-BF16Qwen2.5-14B-InstructGLM-5HunyuanOCRMiniMax-M2.5GLM-4.5-AirSmolLM3-3BQwen3.5-4B-Basegemma-2-9b-itHy3Qwen3.5-122B-A10BNVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16Qwen2.5-Coder-32B-Instructgranite-4.1-3bQwen3.5-397B-A17Bgemma-4-E4BMiniMax-M2.7GLM-5.1MiniCPM-o-4_5Qwen3-Coder-NextQwen3-Coder-30B-A3B-InstructOlmo-3-7B-ThinkQwen3.5-0.8B-BaseDeepSeek-R1-0528-Qwen3-8BQwen3-235B-A22BLaguna-XS-2.1GLM-5.2Olmo-3-7B-Instructgemma-3-27b-itDeepSeek-R1-Distill-Qwen-7BPhi-3.5-mini-instructKimi-K2.6Qwen3.5-9B-BasePhi-4-mini-instructDeepSeek-R1-Distill-Qwen-32Bgemma-3-12b-itphi-4Laguna-S-2.1MiniCPM5-1Bgemma-4-31BKimi-K2-InstructDeepSeek-V3.1DeepSeek-R1-Distill-Qwen-14BKimi-K2.5Qwen3.5-2B-BaseQwen3-VL-235B-A22B-InstructNemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16Devstral-Small-2-24B-Instruct-2512Qwen3-VL-32B-InstructStep-3.5-FlashNVIDIA-Nemotron-Nano-9B-v2Qwen3-VL-30B-A3B-InstructQwen2.5-72B-InstructLlama-3.3-70B-InstructMuse-Glimmer-30BMistral-Nemo-Instruct-2407Mistral-Small-3.2-24B-Instruct-2506gemma-4-E2BMiniCPM-V-4.6Qwen3-Next-80B-A3B-InstructKimi-Linear-48B-A3B-InstructMiniMax-M2EuroLLM-22B-Instruct-2512granite-vision-4.1-4bQwen3-235B-A22B-Instruct-2507MiMo-V2.5Llama-4-Scout-17B-16E-InstructGLM-5.3-Flash-BF16DeepSeek-R1-Distill-Llama-8BQwen3.5-35B-A3B-BaseGLM-4.5Ministral-3-14B-Instruct-2512Ministral-3-8B-Instruct-2512LFM2.5-1.2B-InstructLFM2.5-2.6BMinistral-3-14B-Reasoning-2512Qwen3-VL-8B-ThinkingMiMo-V2.6-Flash-RLGLM-4.6V-FlashMinistral-3-3B-Instruct-2512MiMo-V2.6-Pro-RLCosmos-Reason2-8Bgranite-4.1-8bgranite-4.2-8bgranite-3.3-8b-instructLlama-3.1-70B-Instructgranite-4.1-30bGLM-4.7LFM2.5-230MolmOCR-2-7B-1025Qwen3-VL-2B-Thinkinggemma-4-12BMistral-Medium-3.5-128Bmedgemma-1.5-4b-itMistral-Small-4-119B-2603MiniCPM-V-4.6-ThinkingMiMo-V2-FlashMiniCPM-V-4.7-35B-A3Bgranite-guardian-4.1-8bKimi-K2-ThinkingLFM2.5-350MDeepSeek-R1-Distill-Llama-70BEXAONE-4.5-33BOlmo-3-32B-Think-SFTgemma-4-26B-A4BQwen3-Coder-480B-A35B-InstructAI21-Jamba-Reasoning-3Bgranite-4.2-3bLFM2.5-VL-450Mgranite-4.2-30bQwQ-32BOlmo-3-7B-Instruct-SFTLFM2.5-8B-A1BNVIDIA-Nemotron-3.5-Lightning-30B-A3B-Base-BF16LFM2.5-VL-1.6BLFM2.5-VL-3BMiMo-V2.6-Distill-Qwen-9BAI21-Jamba2-3BQwen3.8-2.4T-A95Bgranite-4.0-350mMiniCPM-V-4.7-1BQwen3-VL-4B-ThinkingHolo2-4Btranslategemma-4b-itd1-3Bcommand-a-plus-05-2026-bf16Olmo-3.1-32B-ThinkStep-3.7-Flashgemma-2-27b-itOlmo-3.1-32B-InstructEuroLLM-9B-Instruct-2512MiMo-V2.5-Profunctiongemma-270m-itOlmo-3-1125-32BMinistral-3-3B-Reasoning-2512UI-Venus-2-9BMellum2-12B-A2.5B-InstructOlmo-3-32B-ThinkLFM2-8B-A1BMellum2-12B-A2.5B-BaseK-EXAONE-236B-A23BLFM2.5-24B-A2BMinistral-3-8B-Reasoning-2512Olmo-3-7B-Instruct-DPOKAT-Coder-V2.5-DevNVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16Trinity-MiniYoutu-LLM-2BHolo-3.1-4BNemotron-3-Content-SafetyMiniMax-M2.1LFM2.5-1.2B-Basegranite-4.0-1bQwen3-VL-32B-ThinkingMiniCPM5-2B-BaseMiMo-V2.6-Flash-MOPDtranslategemma-12b-itSolar-Open-100BOlmo-3-7B-Think-SFTLFM2.5-1.2B-JP-202606Nemotron-Cascade-2-30B-A3BQwen-AgentWorld-35B-A3Btiny-aya-globalLFM2.5-1.2B-ThinkingNorth-Mini-Code-1.0Holo-3.1-0.8BNVIDIA-Nemotron-3-Super-120B-A12B-Base-BF16MiMo-V2.6-Pro-MOPDMiniCPM5-1B-Basetranslategemma-27b-itOlmo-3-7B-Think-DPOBielik-Minitron-7B-v3.0-InstructLFM2.5-2.6B-BaseMiniCPM5-1B-SFTDevstral-2-123B-Instruct-2512UI-Venus-1.5-2BFara1.5-9Bkanana-2-3b-instructMellum2-12B-A2.5B-Thinking-SFTIsing-Calibration-1.5-31B-BF16granite-4.0-h-350mTrinity-Nano-PreviewFara-7BMellum2.1-12B-A2.5B-ThinkingBielik-11B-v3.0-InstructOlmo-3.1-32B-Instruct-DPOGLM-4.6VKimi-Linear-48B-A3B-BaseMellum2-12B-A2.5B-ThinkingLFM2-24B-A2BLFM2-VL-3BOlmo-3-7B-RL-Zero-Mathgranite-4.0-h-1bCosmos-Reason2-32BBitCPM-CANN-3B-unquantizedEuroLLM-22B-2512Olmo-3.1-7B-RL-Zero-Mathtiny-aya-earthNemotron-Orchestrator-8BHolo4-35B-A3BAutoGLM-Phone-9BMistral-Large-3-675B-Instruct-2512WebWorld-32BTrinity-Large-ThinkingFara1.5-27BLing-1TAstaBrief_8BOptiMind-SFTUI-Venus-1.5-8BFrogNano-4B-2609Olmo-3-7B-RL-Zero-IFOlmo-3-7B-RL-Zero-CodeMuse-Spark-Open-816B-A42BApertus-v1.1-1.5BQwen3-Coder-Next-BaseOlmo-3-7B-RL-Zero-Generalrnj-1-instructYoutu-VL-4B-InstructBeam-1.0-501B-A23BHermes-4.3-36BApriel-1.6-15b-ThinkerAprielGuardEuroMoE-2.6B-A0.6B-Instruct-2512Fara1.5-4BHolo4-27BUI-Venus-1.5-30B-A3BNemotron-Cascade-14B-ThinkingQwen3.8-3.6-27B-blendLFM2.5-1.2B-JPApertus-v1.1-1.5B-InstructOlmo-3-32B-Think-DPOLFM2.5-Encoder-350M-DiffusionRiva-Translate-4B-Instruct-v1.1Holo-3.1-9BTrinity-Nano-BaseNemotron-Cascade-8B-ThinkingIntern-S2-397BYoutu-LLM-2B-BaseAutoGLM-Phone-9B-MultilingualOlmo-3-7B-RL-Zero-MixCapRL-Qwen3VL-4BLFM2-2.6B-LongevityUI-Mate-9BKimi-Linear-REAP-35B-A3B-InstructWebWorld-8BAI21-Jamba2-MiniMiMo-Embodied-7BAesCode-8BUserRepGen-4BAesCode-32BContextPilot-14BContextPilot-E4BLFM2-2.6B-TranscriptHolotron4-30B-A3Bkanana-2-3b-baseUI-Mate-27BBaichuan-M3-235BApertus-v1.1-4BNemotron-Cascade-8BINTELLECT-3Ising-Calibration-1-35B-A3BContextPilot-8BLFM2-2.6B-ExpNVIDIA-Nemotron-Labs-3-Elastic-30B-A3B-BF16Nemotron-Terminal-8Btiny-aya-fireHARC-Qwen2.5-7B-InstructWebWorld-14BAgentCPM-ExploreOlmo-3.1-7B-RL-Zero-CodeHARC-Llama-3.1-8B-InstructTrinity-Large-TrueBaseSingGuard-2bX-Reasoner-7BMathForm-8BNousCoder-14BTrinity-Large-BaseNemotron-Terminal-32Btiny-aya-waterHolo-3.1-35B-A3BQwen2.5-VL-7B-Surg-CholecT50Nemotron-Terminal-14BFrogBoss-32B-2510Holo3-35B-A3BUI-Mate-democua-27BTrinity-Mini-Base-Pre-AnnealGLM-4.7-Flash-REAP-23B-A3BTrinity-Mini-BaseSmolLM3-3B-GSM8K-SFTcogito-671b-v2.1GELab-Zero-4B-previewCapRL-Qwen3VL-2BJanusCoderV-7BLFM2.5-VL-1.6B-Extractgranite-guardian-4.0-3b-toxicity-jaQwen3-Coder-REAP-25B-A3BLFM2.5-VL-450M-ExtractJanusCoder-14BStep3-VL-10B-Basegranite-guardian-3.2-8b-factuality-detectionZwZ-4BFrogMini-14B-2510Qwen3-Nemotron-235B-A22B-GenRMTrinity-Nano-Base-Pre-AnnealMellum2-12B-A2.5B-Instruct-SFTKAT-Dev-72B-ExpQwen3-Nemotron-8B-BRRMZwZ-8BGLM-4.5-Air-REAP-82B-A12BRLVR-8B-0926Spatial-SSRL-Qwen3VL-4BJanusCoder-8BArmorOCRSpatial-SSRL-7BHolo2-8BSERA-32BHolo2-30B-A3BZwZ-2BQwen3-Coder-REAP-363B-A35BHolo2-235B-A22BMiniMax-M2.5-REAP-139B-A10BQwen3-Nemotron-14B-BRRMintent-aware-lfqa-llama3-8b-intent-explicitStep-3.5-Flash-REAP-121B-A11BGLM-4.7-REAP-218B-A32Bintent-aware-lfqa-llama3-8b-intent-implicitStep-3.5-Flash-REAP-149B-A11Bintent-aware-lfqa-qwen3-4b-intent-explicitGLM-4.6-REAP-268B-A32BGLM-4.7-REAP-268B-A32Bintent-aware-lfqa-qwen3-8b-intent-explicitintent-aware-lfqa-qwen3-4b-intent-implicitintent-aware-lfqa-qwen3-8b-multiviewintent-aware-lfqa-llama3-8b-multiviewintent-aware-lfqa-qwen3-8b-intent-implicitintent-aware-lfqa-qwen3-4b-baselineZwZ-7Bintent-aware-lfqa-qwen3-8b-baselineintent-aware-lfqa-llama3-8b-baselineintent-aware-lfqa-qwen3-4b-multiviewHiPO-1.7B

能装下 BF16，8,192 token 上下文

每张卡“能装下”所需的可用显存（X）：3.41 GiB

在计算器中打开（可设置多张相同的显卡、其他上下文长度和 KV 缓存类型）

按显存容量浏览

• 8 GB
• 12 GB
• 16 GB
• 24 GB
• 32 GB

最近添加的模型

• Qwen3-VL-8B-Thinking Qwen
• Qwen3-VL-32B-Thinking Qwen
• Laguna-XS-2.1 poolside
• Muse-Glimmer-30B meta-models
• Mellum2-12B-A2.5B-Thinking JetBrains
• Mellum2-12B-A2.5B-Instruct JetBrains
• Qwen3.8-3.6-27B-blend JetBrains
• ZwZ-4B inclusionAI
• granite-4.1-3b ibm-granite
• granite-4.1-30b ibm-granite
• granite-4.1-8b ibm-granite
• Holo4-27B Hcompany