[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"$f13yf1xlidjit2":3,"$f261oz3g6d56u0":53,"$f1afu761dszixy":55,"$fxtncuc9aw2hw":248,"$f8ypa021er9qz":250,"$f16rb8hc37tsmx":252},{"id":4,"name":5,"publisher":6,"paramsB":7,"useCase":8,"licenseStatus":9,"licenseUrl":10,"sourceUrl":11,"descriptionZh":12,"descriptionEn":13,"imageUrl":14,"downloadUrls":15,"stats":18,"variants":23,"derivations":24,"measuredHardware":27,"updatedAt":52},9,"GLM-5.3-Flash","Z.ai",320,"CHAT","OPEN_SOURCE","https:\u002F\u002Fhuggingface.co\u002Fzai-org\u002FGLM-5.3-Flash\u002Fblob\u002Fmain\u002FLICENSE","https:\u002F\u002Fhuggingface.co\u002Fzai-org\u002FGLM-5.3-Flash","Z.ai（原智谱）2026-08-26 发布的开权重 MoE 多模态模型：320B 总参数 \u002F 18B 激活，1M 上下文，MIT 许可。原生支持文本、图片、视频与文档输入，支持 SGLang \u002F vLLM 本地部署。","Z.ai's open-weight MoE multimodal model released 2026-08-26: 320B total parameters with 18B activated, 1M context, MIT licensed. Takes text, images, video and documents, and runs locally on SGLang or vLLM.","\u002Fuploads\u002Fmodels\u002Fm9-491ca77152b1.png",[16],{"url":11,"label":17},"HuggingFace",{"recordCount":19,"variantCount":20,"hardwareCount":21,"unlinkedRecordCount":19,"evidence":22},29,0,8,{"L0":19,"L1":20,"L2":20},[],{"parents":25,"children":26},[],[],[28,32,35,38,41,45,48,50],{"name":29,"vramGb":30,"recordCount":31},"NVIDIA RTX 5070 Ti",16,7,{"name":33,"vramGb":30,"recordCount":34},"NVIDIA RTX 4080 Super",6,{"name":36,"vramGb":37,"recordCount":34},"NVIDIA RTX 4090",24,{"name":39,"vramGb":40,"recordCount":34},"NVIDIA RTX 5090",32,{"name":42,"vramGb":43,"recordCount":44},"2x NVIDIA RTX 5090",64,1,{"name":46,"vramGb":47,"recordCount":44},"NVIDIA RTX 3060 12GB",12,{"name":49,"vramGb":37,"recordCount":44},"NVIDIA RTX 3090",{"name":51,"vramGb":47,"recordCount":44},"NVIDIA RTX 5070","2026-09-24T12:55:08.720Z",{"favorited":54},false,{"items":56,"total":19,"page":44,"pageSize":247},[57,74,81,88,96,102,108,114,120,125,131,137,143,150,156,162,168,174,177,183,186,189,194,201,207,216,223,229,235],{"modelId":4,"modelName":5,"hardwareId":58,"hardwareName":33,"frameworkId":59,"frameworkName":60,"quantization":61,"hardwareVramGb":30,"decodeTps":62,"prefillTps":63,"vramGb":30,"genSeconds":64,"modelUseCase":8,"ttftS":65,"ttfbGb":64,"mtpAcceptanceRate":64,"powerW":64,"sampleCount":44,"reproduceCount":44,"sourceTypes":66,"evidenceLevel":68,"updatedAt":69,"latestFrameworkVersion":70,"latestOs":71,"latestCudaVersion":72,"latestFlashAttention":73,"latestOutputSpec":64},20,5,"SGLang","Q4_K_M",3.1,23.3,null,4.29,[67],"OTHER","L0","2026-10-06T12:06:15.424Z","0.4.4","Linux","12.6",true,{"modelId":4,"modelName":5,"hardwareId":58,"hardwareName":33,"frameworkId":59,"frameworkName":60,"quantization":75,"hardwareVramGb":30,"decodeTps":76,"prefillTps":77,"vramGb":30,"genSeconds":64,"modelUseCase":8,"ttftS":78,"ttfbGb":64,"mtpAcceptanceRate":64,"powerW":64,"sampleCount":44,"reproduceCount":44,"sourceTypes":79,"evidenceLevel":68,"updatedAt":80,"latestFrameworkVersion":70,"latestOs":71,"latestCudaVersion":72,"latestFlashAttention":73,"latestOutputSpec":64},"IQ3_S",4.7,35.3,2.83,[67],"2026-10-06T12:06:15.416Z",{"modelId":4,"modelName":5,"hardwareId":58,"hardwareName":33,"frameworkId":59,"frameworkName":60,"quantization":82,"hardwareVramGb":30,"decodeTps":83,"prefillTps":84,"vramGb":30,"genSeconds":64,"modelUseCase":8,"ttftS":85,"ttfbGb":64,"mtpAcceptanceRate":64,"powerW":64,"sampleCount":44,"reproduceCount":44,"sourceTypes":86,"evidenceLevel":68,"updatedAt":87,"latestFrameworkVersion":70,"latestOs":71,"latestCudaVersion":72,"latestFlashAttention":73,"latestOutputSpec":64},"IQ4_XS",3.5,26.3,3.8,[67],"2026-10-06T12:06:15.407Z",{"modelId":4,"modelName":5,"hardwareId":58,"hardwareName":33,"frameworkId":89,"frameworkName":90,"quantization":61,"hardwareVramGb":30,"decodeTps":91,"prefillTps":37,"vramGb":30,"genSeconds":64,"modelUseCase":8,"ttftS":92,"ttfbGb":64,"mtpAcceptanceRate":64,"powerW":64,"sampleCount":44,"reproduceCount":44,"sourceTypes":93,"evidenceLevel":68,"updatedAt":94,"latestFrameworkVersion":95,"latestOs":71,"latestCudaVersion":72,"latestFlashAttention":73,"latestOutputSpec":64},2,"vLLM",3.2,4.17,[67],"2026-10-06T12:06:15.398Z","0.7.3",{"modelId":4,"modelName":5,"hardwareId":58,"hardwareName":33,"frameworkId":89,"frameworkName":90,"quantization":75,"hardwareVramGb":30,"decodeTps":97,"prefillTps":98,"vramGb":30,"genSeconds":64,"modelUseCase":8,"ttftS":99,"ttfbGb":64,"mtpAcceptanceRate":64,"powerW":64,"sampleCount":44,"reproduceCount":44,"sourceTypes":100,"evidenceLevel":68,"updatedAt":101,"latestFrameworkVersion":95,"latestOs":71,"latestCudaVersion":72,"latestFlashAttention":73,"latestOutputSpec":64},4.8,36,2.78,[67],"2026-10-06T12:06:15.387Z",{"modelId":4,"modelName":5,"hardwareId":58,"hardwareName":33,"frameworkId":89,"frameworkName":90,"quantization":82,"hardwareVramGb":30,"decodeTps":103,"prefillTps":104,"vramGb":30,"genSeconds":64,"modelUseCase":8,"ttftS":105,"ttfbGb":64,"mtpAcceptanceRate":64,"powerW":64,"sampleCount":44,"reproduceCount":44,"sourceTypes":106,"evidenceLevel":68,"updatedAt":107,"latestFrameworkVersion":95,"latestOs":71,"latestCudaVersion":72,"latestFlashAttention":73,"latestOutputSpec":64},3.6,27,3.7,[67],"2026-10-06T12:06:15.378Z",{"modelId":4,"modelName":5,"hardwareId":109,"hardwareName":29,"frameworkId":59,"frameworkName":60,"quantization":61,"hardwareVramGb":30,"decodeTps":85,"prefillTps":110,"vramGb":30,"genSeconds":64,"modelUseCase":8,"ttftS":111,"ttfbGb":64,"mtpAcceptanceRate":64,"powerW":64,"sampleCount":44,"reproduceCount":44,"sourceTypes":112,"evidenceLevel":68,"updatedAt":113,"latestFrameworkVersion":70,"latestOs":71,"latestCudaVersion":72,"latestFlashAttention":73,"latestOutputSpec":64},17,28.5,3.51,[67],"2026-10-06T12:06:15.366Z",{"modelId":4,"modelName":5,"hardwareId":109,"hardwareName":29,"frameworkId":59,"frameworkName":60,"quantization":75,"hardwareVramGb":30,"decodeTps":115,"prefillTps":116,"vramGb":30,"genSeconds":64,"modelUseCase":8,"ttftS":117,"ttfbGb":64,"mtpAcceptanceRate":64,"powerW":64,"sampleCount":44,"reproduceCount":44,"sourceTypes":118,"evidenceLevel":68,"updatedAt":119,"latestFrameworkVersion":70,"latestOs":71,"latestCudaVersion":72,"latestFlashAttention":73,"latestOutputSpec":64},5.7,42.8,2.34,[67],"2026-10-06T12:06:15.356Z",{"modelId":4,"modelName":5,"hardwareId":109,"hardwareName":29,"frameworkId":59,"frameworkName":60,"quantization":82,"hardwareVramGb":30,"decodeTps":121,"prefillTps":122,"vramGb":30,"genSeconds":64,"modelUseCase":8,"ttftS":62,"ttfbGb":64,"mtpAcceptanceRate":64,"powerW":64,"sampleCount":44,"reproduceCount":44,"sourceTypes":123,"evidenceLevel":68,"updatedAt":124,"latestFrameworkVersion":70,"latestOs":71,"latestCudaVersion":72,"latestFlashAttention":73,"latestOutputSpec":64},4.3,32.3,[67],"2026-10-06T12:06:15.348Z",{"modelId":4,"modelName":5,"hardwareId":109,"hardwareName":29,"frameworkId":89,"frameworkName":90,"quantization":61,"hardwareVramGb":30,"decodeTps":126,"prefillTps":127,"vramGb":30,"genSeconds":64,"modelUseCase":8,"ttftS":128,"ttfbGb":64,"mtpAcceptanceRate":64,"powerW":64,"sampleCount":44,"reproduceCount":44,"sourceTypes":129,"evidenceLevel":68,"updatedAt":130,"latestFrameworkVersion":95,"latestOs":71,"latestCudaVersion":72,"latestFlashAttention":73,"latestOutputSpec":64},3.9,29.3,3.41,[67],"2026-10-06T12:06:15.338Z",{"modelId":4,"modelName":5,"hardwareId":109,"hardwareName":29,"frameworkId":89,"frameworkName":90,"quantization":75,"hardwareVramGb":30,"decodeTps":132,"prefillTps":133,"vramGb":30,"genSeconds":64,"modelUseCase":8,"ttftS":134,"ttfbGb":64,"mtpAcceptanceRate":64,"powerW":64,"sampleCount":44,"reproduceCount":44,"sourceTypes":135,"evidenceLevel":68,"updatedAt":136,"latestFrameworkVersion":95,"latestOs":71,"latestCudaVersion":72,"latestFlashAttention":73,"latestOutputSpec":64},5.9,44.3,2.26,[67],"2026-10-06T12:06:15.328Z",{"modelId":4,"modelName":5,"hardwareId":109,"hardwareName":29,"frameworkId":89,"frameworkName":90,"quantization":82,"hardwareVramGb":30,"decodeTps":138,"prefillTps":139,"vramGb":30,"genSeconds":64,"modelUseCase":8,"ttftS":140,"ttfbGb":64,"mtpAcceptanceRate":64,"powerW":64,"sampleCount":44,"reproduceCount":44,"sourceTypes":141,"evidenceLevel":68,"updatedAt":142,"latestFrameworkVersion":95,"latestOs":71,"latestCudaVersion":72,"latestFlashAttention":73,"latestOutputSpec":64},4.4,33,3.03,[67],"2026-10-06T12:06:15.319Z",{"modelId":4,"modelName":5,"hardwareId":144,"hardwareName":39,"frameworkId":59,"frameworkName":60,"quantization":61,"hardwareVramGb":40,"decodeTps":145,"prefillTps":146,"vramGb":40,"genSeconds":64,"modelUseCase":8,"ttftS":147,"ttfbGb":64,"mtpAcceptanceRate":64,"powerW":64,"sampleCount":44,"reproduceCount":44,"sourceTypes":148,"evidenceLevel":68,"updatedAt":149,"latestFrameworkVersion":70,"latestOs":71,"latestCudaVersion":72,"latestFlashAttention":73,"latestOutputSpec":64},13,7.6,57,1.75,[67],"2026-10-06T12:06:15.310Z",{"modelId":4,"modelName":5,"hardwareId":144,"hardwareName":39,"frameworkId":59,"frameworkName":60,"quantization":82,"hardwareVramGb":40,"decodeTps":151,"prefillTps":152,"vramGb":40,"genSeconds":64,"modelUseCase":8,"ttftS":153,"ttfbGb":64,"mtpAcceptanceRate":64,"powerW":64,"sampleCount":44,"reproduceCount":44,"sourceTypes":154,"evidenceLevel":68,"updatedAt":155,"latestFrameworkVersion":70,"latestOs":71,"latestCudaVersion":72,"latestFlashAttention":73,"latestOutputSpec":64},8.5,63.8,1.57,[67],"2026-10-06T12:06:15.300Z",{"modelId":4,"modelName":5,"hardwareId":144,"hardwareName":39,"frameworkId":89,"frameworkName":90,"quantization":61,"hardwareVramGb":40,"decodeTps":157,"prefillTps":158,"vramGb":40,"genSeconds":64,"modelUseCase":8,"ttftS":159,"ttfbGb":64,"mtpAcceptanceRate":64,"powerW":64,"sampleCount":44,"reproduceCount":44,"sourceTypes":160,"evidenceLevel":68,"updatedAt":161,"latestFrameworkVersion":95,"latestOs":71,"latestCudaVersion":72,"latestFlashAttention":73,"latestOutputSpec":64},7.8,58.5,1.71,[67],"2026-10-06T12:06:15.293Z",{"modelId":4,"modelName":5,"hardwareId":144,"hardwareName":39,"frameworkId":89,"frameworkName":90,"quantization":75,"hardwareVramGb":40,"decodeTps":163,"prefillTps":164,"vramGb":40,"genSeconds":64,"modelUseCase":8,"ttftS":165,"ttfbGb":64,"mtpAcceptanceRate":64,"powerW":64,"sampleCount":44,"reproduceCount":44,"sourceTypes":166,"evidenceLevel":68,"updatedAt":167,"latestFrameworkVersion":95,"latestOs":71,"latestCudaVersion":72,"latestFlashAttention":73,"latestOutputSpec":64},11.8,88.5,1.13,[67],"2026-10-06T12:06:15.286Z",{"modelId":4,"modelName":5,"hardwareId":144,"hardwareName":39,"frameworkId":89,"frameworkName":90,"quantization":82,"hardwareVramGb":40,"decodeTps":169,"prefillTps":170,"vramGb":40,"genSeconds":64,"modelUseCase":8,"ttftS":171,"ttfbGb":64,"mtpAcceptanceRate":64,"powerW":64,"sampleCount":44,"reproduceCount":44,"sourceTypes":172,"evidenceLevel":68,"updatedAt":173,"latestFrameworkVersion":95,"latestOs":71,"latestCudaVersion":72,"latestFlashAttention":73,"latestOutputSpec":64},8.8,66,1.52,[67],"2026-10-06T12:06:15.277Z",{"modelId":4,"modelName":5,"hardwareId":44,"hardwareName":36,"frameworkId":59,"frameworkName":60,"quantization":61,"hardwareVramGb":37,"decodeTps":121,"prefillTps":122,"vramGb":37,"genSeconds":64,"modelUseCase":8,"ttftS":62,"ttfbGb":64,"mtpAcceptanceRate":64,"powerW":64,"sampleCount":44,"reproduceCount":44,"sourceTypes":175,"evidenceLevel":68,"updatedAt":176,"latestFrameworkVersion":70,"latestOs":71,"latestCudaVersion":72,"latestFlashAttention":73,"latestOutputSpec":64},[67],"2026-10-06T12:06:15.267Z",{"modelId":4,"modelName":5,"hardwareId":44,"hardwareName":36,"frameworkId":59,"frameworkName":60,"quantization":75,"hardwareVramGb":37,"decodeTps":178,"prefillTps":179,"vramGb":37,"genSeconds":64,"modelUseCase":8,"ttftS":180,"ttfbGb":64,"mtpAcceptanceRate":64,"powerW":64,"sampleCount":44,"reproduceCount":44,"sourceTypes":181,"evidenceLevel":68,"updatedAt":182,"latestFrameworkVersion":70,"latestOs":71,"latestCudaVersion":72,"latestFlashAttention":73,"latestOutputSpec":64},6.4,48,2.08,[67],"2026-10-06T12:06:15.258Z",{"modelId":4,"modelName":5,"hardwareId":44,"hardwareName":36,"frameworkId":59,"frameworkName":60,"quantization":82,"hardwareVramGb":37,"decodeTps":97,"prefillTps":98,"vramGb":37,"genSeconds":64,"modelUseCase":8,"ttftS":99,"ttfbGb":64,"mtpAcceptanceRate":64,"powerW":64,"sampleCount":44,"reproduceCount":44,"sourceTypes":184,"evidenceLevel":68,"updatedAt":185,"latestFrameworkVersion":70,"latestOs":71,"latestCudaVersion":72,"latestFlashAttention":73,"latestOutputSpec":64},[67],"2026-10-06T12:06:15.249Z",{"modelId":4,"modelName":5,"hardwareId":44,"hardwareName":36,"frameworkId":89,"frameworkName":90,"quantization":61,"hardwareVramGb":37,"decodeTps":138,"prefillTps":139,"vramGb":37,"genSeconds":64,"modelUseCase":8,"ttftS":140,"ttfbGb":64,"mtpAcceptanceRate":64,"powerW":64,"sampleCount":44,"reproduceCount":44,"sourceTypes":187,"evidenceLevel":68,"updatedAt":188,"latestFrameworkVersion":95,"latestOs":71,"latestCudaVersion":72,"latestFlashAttention":73,"latestOutputSpec":64},[67],"2026-10-06T12:06:15.241Z",{"modelId":4,"modelName":5,"hardwareId":44,"hardwareName":36,"frameworkId":89,"frameworkName":90,"quantization":82,"hardwareVramGb":37,"decodeTps":59,"prefillTps":190,"vramGb":37,"genSeconds":64,"modelUseCase":8,"ttftS":191,"ttfbGb":64,"mtpAcceptanceRate":64,"powerW":64,"sampleCount":44,"reproduceCount":44,"sourceTypes":192,"evidenceLevel":68,"updatedAt":193,"latestFrameworkVersion":95,"latestOs":71,"latestCudaVersion":72,"latestFlashAttention":73,"latestOutputSpec":64},37.5,2.67,[67],"2026-10-06T12:06:15.231Z",{"modelId":4,"modelName":5,"hardwareId":195,"hardwareName":46,"frameworkId":44,"frameworkName":196,"quantization":197,"hardwareVramGb":47,"decodeTps":62,"prefillTps":63,"vramGb":47,"genSeconds":64,"modelUseCase":8,"ttftS":65,"ttfbGb":64,"mtpAcceptanceRate":64,"powerW":64,"sampleCount":44,"reproduceCount":44,"sourceTypes":198,"evidenceLevel":68,"updatedAt":199,"latestFrameworkVersion":200,"latestOs":71,"latestCudaVersion":72,"latestFlashAttention":73,"latestOutputSpec":64},15,"llama.cpp","IQ2_XXS",[67],"2026-10-06T03:39:17.951Z","b4200",{"modelId":4,"modelName":5,"hardwareId":89,"hardwareName":49,"frameworkId":44,"frameworkName":196,"quantization":75,"hardwareVramGb":37,"decodeTps":202,"prefillTps":203,"vramGb":37,"genSeconds":64,"modelUseCase":8,"ttftS":204,"ttfbGb":64,"mtpAcceptanceRate":64,"powerW":64,"sampleCount":44,"reproduceCount":44,"sourceTypes":205,"evidenceLevel":68,"updatedAt":206,"latestFrameworkVersion":200,"latestOs":71,"latestCudaVersion":72,"latestFlashAttention":73,"latestOutputSpec":64},5.6,42,2.38,[67],"2026-10-06T03:39:17.940Z",{"modelId":4,"modelName":5,"hardwareId":109,"hardwareName":29,"frameworkId":4,"frameworkName":208,"quantization":209,"hardwareVramGb":30,"decodeTps":210,"prefillTps":211,"vramGb":30,"genSeconds":64,"modelUseCase":8,"ttftS":212,"ttfbGb":64,"mtpAcceptanceRate":64,"powerW":64,"sampleCount":44,"reproduceCount":44,"sourceTypes":213,"evidenceLevel":68,"updatedAt":214,"latestFrameworkVersion":215,"latestOs":71,"latestCudaVersion":72,"latestFlashAttention":73,"latestOutputSpec":64},"Strata","IQ2_XS",8.3,62.3,1.61,[67],"2026-10-06T03:39:17.929Z","0.5.2",{"modelId":4,"modelName":5,"hardwareId":217,"hardwareName":51,"frameworkId":44,"frameworkName":196,"quantization":209,"hardwareVramGb":47,"decodeTps":218,"prefillTps":219,"vramGb":47,"genSeconds":64,"modelUseCase":8,"ttftS":220,"ttfbGb":64,"mtpAcceptanceRate":64,"powerW":64,"sampleCount":44,"reproduceCount":44,"sourceTypes":221,"evidenceLevel":68,"updatedAt":222,"latestFrameworkVersion":200,"latestOs":71,"latestCudaVersion":72,"latestFlashAttention":73,"latestOutputSpec":64},18,4.9,36.8,2.72,[67],"2026-10-06T03:39:17.920Z",{"modelId":4,"modelName":5,"hardwareId":144,"hardwareName":39,"frameworkId":59,"frameworkName":60,"quantization":75,"hardwareVramGb":40,"decodeTps":224,"prefillTps":225,"vramGb":40,"genSeconds":64,"modelUseCase":8,"ttftS":226,"ttfbGb":64,"mtpAcceptanceRate":64,"powerW":64,"sampleCount":44,"reproduceCount":44,"sourceTypes":227,"evidenceLevel":68,"updatedAt":228,"latestFrameworkVersion":70,"latestOs":71,"latestCudaVersion":72,"latestFlashAttention":73,"latestOutputSpec":64},11.3,84.8,1.18,[67],"2026-10-06T03:39:17.911Z",{"modelId":4,"modelName":5,"hardwareId":44,"hardwareName":36,"frameworkId":89,"frameworkName":90,"quantization":75,"hardwareVramGb":37,"decodeTps":230,"prefillTps":231,"vramGb":37,"genSeconds":64,"modelUseCase":8,"ttftS":232,"ttfbGb":64,"mtpAcceptanceRate":64,"powerW":64,"sampleCount":44,"reproduceCount":44,"sourceTypes":233,"evidenceLevel":68,"updatedAt":234,"latestFrameworkVersion":95,"latestOs":71,"latestCudaVersion":72,"latestFlashAttention":73,"latestOutputSpec":64},6.6,49.5,2.02,[67],"2026-10-06T03:39:17.902Z",{"modelId":4,"modelName":5,"hardwareId":34,"hardwareName":42,"frameworkId":44,"frameworkName":196,"quantization":236,"hardwareVramGb":43,"decodeTps":237,"prefillTps":238,"vramGb":239,"genSeconds":64,"modelUseCase":8,"ttftS":240,"ttfbGb":64,"mtpAcceptanceRate":64,"powerW":64,"sampleCount":44,"reproduceCount":44,"sourceTypes":241,"evidenceLevel":68,"updatedAt":243,"latestFrameworkVersion":244,"latestOs":245,"latestCudaVersion":246,"latestFlashAttention":64,"latestOutputSpec":64},"UD-IQ4_XS",21.96,421.07,60.34,0.596,[242],"GITHUB","2026-09-27T11:51:40.843Z","moecachev1.6-rc0","Ubuntu 24.04.4 LTS","13.3",50,{"items":249,"total":20,"page":44,"pageSize":58},[],{"items":251},[],{"items":253,"total":89,"page":44,"pageSize":300},[254,274],{"slug":255,"category":256,"publishedAt":257,"titleZh":258,"titleEn":259,"summaryZh":260,"summaryEn":261,"models":262,"hardwares":273},"minimax-m31-flash-2026-10-20","NEWS","2026-10-20T10:00:00.000Z","MiniMax-M3.1-Flash 发布：API 先行、权重后置，在线 3 倍速、成本降 70%","MiniMax-M3.1-Flash ships: API first, weights later, 3x faster online at 70% lower cost","10 月 19 日 MiniMax 上线 M3.1-Flash（Flash 系列新旗舰，API 先行）。官方口径：在线推理约 3 倍速、任务成本降 70%（$0.15 \u002F $0.60 \u002F $0.015 每百万 token），输出 200+ tok\u002Fs。匿名声测代号 Space Bunny，已处理 200 万亿 token。权重稍后放出——站内按 RESTRICTED \u002F 参数未公布口径收录，零实测记录。","MiniMax shipped M3.1-Flash on Oct 19, the new Flash-series flagship, API first. Official claims: roughly 3x faster online inference, 70% lower task cost ($0.15 \u002F $0.60 \u002F $0.015 per 1M tokens), 200+ tok\u002Fs output. It ran anonymously as Space Bunny, processing 200 trillion tokens. Weights come later — logged as RESTRICTED with unpublished params, zero on-site runs.",[263,264,267,270],{"id":4,"name":5},{"id":265,"name":266},11,"Qwen3.8-Flash-Next",{"id":268,"name":269},22,"MiniMax-M3",{"id":271,"name":272},91,"MiniMax-M3.1-Flash",[],{"slug":275,"category":256,"publishedAt":276,"titleZh":277,"titleEn":278,"summaryZh":279,"summaryEn":280,"models":281,"hardwares":298},"open-source-llm-5way-2026-10-09","2026-10-09T16:00:00.000Z","开源大模型 5 家横评：Qwen 3.x \u002F DeepSeek V4.1 \u002F GLM-5.3 \u002F Llama 4.5 \u002F Gemma 4","Open-source LLM 5-way comparison: Qwen 3.x \u002F DeepSeek V4.1 \u002F GLM-5.3 \u002F Llama 4.5 \u002F Gemma 4","5 家开源旗舰横评，基于本地 DB RTX 4090 实测：Qwen 3.5 27B 91 tok\u002Fs、Gemma 4 12B 89 tok\u002Fs、Qwen 3.8 Flash 180B 12 tok\u002Fs、DeepSeek V4.1 763B 3 tok\u002Fs。27B 是单卡甜蜜点。","5-way open-source flagship comparison based on local DB RTX 4090 measurements: Qwen 3.5 27B 91 tok\u002Fs, Gemma 4 12B 89 tok\u002Fs, Qwen 3.8 Flash 180B 12 tok\u002Fs, DeepSeek V4.1 763B 3 tok\u002Fs. 27B is the sweet spot.",[282,284,285,286,288,290,292,295],{"id":31,"name":283},"DeepSeek-V4-Flash",{"id":4,"name":5},{"id":265,"name":266},{"id":217,"name":287},"DeepSeek-V4.1-Flash",{"id":58,"name":289},"GLM-5.3",{"id":247,"name":291},"Qwen3.5-27B",{"id":293,"name":294},63,"gemma-4-31B-it",{"id":296,"name":297},65,"gemma-4-12B-it",[299],{"id":44,"name":36},3]