[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill-together-ai-together-gpu-clusters":3,"mdc--opqpud-key":37,"related-org-together-ai-together-gpu-clusters":692,"related-repo-together-ai-together-gpu-clusters":864},{"slug":4,"name":4,"fn":5,"description":6,"org":7,"tags":12,"stars":26,"repoUrl":27,"updatedAt":28,"license":29,"forks":30,"topics":31,"repo":32,"sourceUrl":35,"mdContent":36},"together-gpu-clusters","orchestrate GPU clusters on Together AI","On-demand and reserved GPU clusters (H100, H200, B200) on Together AI with Kubernetes or Slurm orchestration, shared storage, credential management, and cluster scaling for ML and HPC jobs. Reach for it when the user needs multi-node compute or infrastructure control rather than a managed model endpoint.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},"together-ai","Together AI","https:\u002F\u002Fpexgzepcugksgbtrxkhf.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Forg-logos\u002Ftogether-ai.jpg","togethercomputer",[13,17,20,23],{"name":14,"slug":15,"type":16},"Machine Learning","machine-learning","tag",{"name":18,"slug":19,"type":16},"Cloud","cloud",{"name":21,"slug":22,"type":16},"AI Infrastructure","ai-infrastructure",{"name":24,"slug":25,"type":16},"Kubernetes","kubernetes",31,"https:\u002F\u002Fgithub.com\u002Ftogethercomputer\u002Fskills","2026-07-26T05:49:05.252646",null,4,[],{"repoUrl":27,"stars":26,"forks":30,"topics":33,"description":34},[],"Skills to help your coding agents use Together AI products.","https:\u002F\u002Fgithub.com\u002Ftogethercomputer\u002Fskills\u002Ftree\u002FHEAD\u002Fskills\u002Ftogether-gpu-clusters","---\nname: together-gpu-clusters\ndescription: \"On-demand and reserved GPU clusters (H100, H200, B200) on Together AI with Kubernetes or Slurm orchestration, shared storage, credential management, and cluster scaling for ML and HPC jobs. Reach for it when the user needs multi-node compute or infrastructure control rather than a managed model endpoint.\"\n---\n\n# Together GPU Clusters\n\n## Overview\n\nUse Together AI GPU clusters when the user needs infrastructure control instead of a managed\ninference product.\n\nTypical fits:\n\n- distributed training\n- multi-node inference\n- HPC or Slurm workloads\n- custom Kubernetes jobs\n- attached shared storage and cluster lifecycle management\n\n## When This Skill Wins\n\n- Provision a cluster and manage it over time\n- Choose between on-demand and reserved capacity\n- Choose Kubernetes or Slurm as the orchestration layer\n- Manage shared volumes and credentials\n- Scale up, scale down, or troubleshoot node health\n\n## Hand Off To Another Skill\n\n- Use `together-dedicated-model-inference` for managed single-model hosting\n- Use `together-dedicated-containers` for containerized inference without owning the full cluster\n- Use `together-sandboxes` for short-lived remote Python execution\n- Use `together-fine-tuning` for managed training jobs instead of raw cluster operations\n\n## Quick Routing\n\n- **Cluster creation, scaling, credentials, deletion**\n  - Start with [scripts\u002Fmanage_cluster.py](scripts\u002Fmanage_cluster.py) or [scripts\u002Fmanage_cluster.ts](scripts\u002Fmanage_cluster.ts)\n  - Read [references\u002Fapi-reference.md](references\u002Fapi-reference.md)\n- **Shared storage lifecycle**\n  - Use [scripts\u002Fmanage_storage.py](scripts\u002Fmanage_storage.py)\n  - Read [references\u002Fapi-reference.md](references\u002Fapi-reference.md)\n- **Kubernetes vs Slurm operations**\n  - Read [references\u002Fcluster-management.md](references\u002Fcluster-management.md)\n- **Troubleshooting node health, PVCs, or scheduling**\n  - Read [references\u002Fcluster-management.md](references\u002Fcluster-management.md)\n- **Together CLI workflows**\n  - Read [references\u002Fcli.md](references\u002Fcli.md)\n\n## Workflow\n\n1. Decide whether the workload really needs cluster-level control.\n2. Choose on-demand vs reserved billing based on run duration and baseline utilization.\n3. Choose Kubernetes vs Slurm based on orchestration requirements and team tooling.\n4. Select region, GPU type, driver version, and shared storage plan.\n5. Provision first, then layer in access credentials, workload deployment, scaling, and health checks.\n\n## High-Signal Rules\n\n- Python scripts require the Together v2 SDK (`together>=2.0.0`). If the user is on an older version, they must upgrade first: `uv pip install --upgrade \"together>=2.0.0\"`.\n- Prefer managed products unless the user explicitly needs raw infrastructure control.\n- Treat storage lifecycle separately from cluster lifecycle; volumes can outlive clusters.\n- When creating a cluster with new shared storage, prefer inline `shared_volume` over creating a volume separately and attaching via `volume_id`. Separately created volumes may land in a different datacenter partition than the cluster, causing a \"does not exist in the datacenter\" error even when the volume shows as available.\n- GPU stock-outs (409 \"Out of stock\") are common. Always call `list_regions()` first and be prepared to try multiple regions.\n- The API requires `cuda_version` and `nvidia_driver_version` as separate fields in addition to the combined `driver_version` string. Pass them via `extra_body` in the Python SDK.\n- Credentials retrieval is part of provisioning. Do not stop at cluster creation if the user needs to run workloads immediately.\n- Slurm and Kubernetes operational patterns differ materially; read the cluster-management reference before improvising.\n- For repeated cluster operations, start from the scripts instead of rebuilding request shapes.\n- Slurm startup scripts (worker\u002Flogin init, worker\u002Fcontroller prolog and epilog, extra `slurm.conf`) are **Slinky v1.0 only**. A non-zero exit from a worker prolog or epilog drains the node, and calling Slurm commands (`squeue`, `scontrol`, `sacctmgr`) inside any prolog\u002Fepilog can deadlock the scheduler.\n\n## Resource Map\n\n- **Cluster API reference**: [references\u002Fapi-reference.md](references\u002Fapi-reference.md)\n- **Operational guide**: [references\u002Fcluster-management.md](references\u002Fcluster-management.md)\n- **Operational troubleshooting**: [references\u002Fcluster-management.md](references\u002Fcluster-management.md)\n- **CLI guide**: [references\u002Fcli.md](references\u002Fcli.md)\n- **Python cluster management**: [scripts\u002Fmanage_cluster.py](scripts\u002Fmanage_cluster.py)\n- **TypeScript cluster management**: [scripts\u002Fmanage_cluster.ts](scripts\u002Fmanage_cluster.ts)\n- **Python storage management**: [scripts\u002Fmanage_storage.py](scripts\u002Fmanage_storage.py)\n\n## Official Docs\n\n- [GPU Clusters Overview](https:\u002F\u002Fdocs.together.ai\u002Fdocs\u002Fgpu-clusters-overview)\n- [GPU Clusters Quickstart](https:\u002F\u002Fdocs.together.ai\u002Fdocs\u002Fgpu-clusters-quickstart)\n- [Clusters API](https:\u002F\u002Fdocs.together.ai\u002Freference\u002Fclusters-create)\n- [Slurm Startup Scripts](https:\u002F\u002Fdocs.together.ai\u002Fdocs\u002Fslurm-startup-scripts)\n- [Instant GPU Clusters](https:\u002F\u002Fwww.together.ai\u002Finstant-gpu-clusters)\n",{"data":38,"body":39},{"name":4,"description":6},{"type":40,"children":41},"root",[42,50,57,63,68,98,104,132,138,191,197,327,333,362,368,531,537,632,638],{"type":43,"tag":44,"props":45,"children":46},"element","h1",{"id":4},[47],{"type":48,"value":49},"text","Together GPU Clusters",{"type":43,"tag":51,"props":52,"children":54},"h2",{"id":53},"overview",[55],{"type":48,"value":56},"Overview",{"type":43,"tag":58,"props":59,"children":60},"p",{},[61],{"type":48,"value":62},"Use Together AI GPU clusters when the user needs infrastructure control instead of a managed\ninference product.",{"type":43,"tag":58,"props":64,"children":65},{},[66],{"type":48,"value":67},"Typical fits:",{"type":43,"tag":69,"props":70,"children":71},"ul",{},[72,78,83,88,93],{"type":43,"tag":73,"props":74,"children":75},"li",{},[76],{"type":48,"value":77},"distributed training",{"type":43,"tag":73,"props":79,"children":80},{},[81],{"type":48,"value":82},"multi-node inference",{"type":43,"tag":73,"props":84,"children":85},{},[86],{"type":48,"value":87},"HPC or Slurm workloads",{"type":43,"tag":73,"props":89,"children":90},{},[91],{"type":48,"value":92},"custom Kubernetes jobs",{"type":43,"tag":73,"props":94,"children":95},{},[96],{"type":48,"value":97},"attached shared storage and cluster lifecycle management",{"type":43,"tag":51,"props":99,"children":101},{"id":100},"when-this-skill-wins",[102],{"type":48,"value":103},"When This Skill Wins",{"type":43,"tag":69,"props":105,"children":106},{},[107,112,117,122,127],{"type":43,"tag":73,"props":108,"children":109},{},[110],{"type":48,"value":111},"Provision a cluster and manage it over time",{"type":43,"tag":73,"props":113,"children":114},{},[115],{"type":48,"value":116},"Choose between on-demand and reserved capacity",{"type":43,"tag":73,"props":118,"children":119},{},[120],{"type":48,"value":121},"Choose Kubernetes or Slurm as the orchestration layer",{"type":43,"tag":73,"props":123,"children":124},{},[125],{"type":48,"value":126},"Manage shared volumes and credentials",{"type":43,"tag":73,"props":128,"children":129},{},[130],{"type":48,"value":131},"Scale up, scale down, or troubleshoot node health",{"type":43,"tag":51,"props":133,"children":135},{"id":134},"hand-off-to-another-skill",[136],{"type":48,"value":137},"Hand Off To Another Skill",{"type":43,"tag":69,"props":139,"children":140},{},[141,155,167,179],{"type":43,"tag":73,"props":142,"children":143},{},[144,146,153],{"type":48,"value":145},"Use ",{"type":43,"tag":147,"props":148,"children":150},"code",{"className":149},[],[151],{"type":48,"value":152},"together-dedicated-model-inference",{"type":48,"value":154}," for managed single-model hosting",{"type":43,"tag":73,"props":156,"children":157},{},[158,159,165],{"type":48,"value":145},{"type":43,"tag":147,"props":160,"children":162},{"className":161},[],[163],{"type":48,"value":164},"together-dedicated-containers",{"type":48,"value":166}," for containerized inference without owning the full cluster",{"type":43,"tag":73,"props":168,"children":169},{},[170,171,177],{"type":48,"value":145},{"type":43,"tag":147,"props":172,"children":174},{"className":173},[],[175],{"type":48,"value":176},"together-sandboxes",{"type":48,"value":178}," for short-lived remote Python execution",{"type":43,"tag":73,"props":180,"children":181},{},[182,183,189],{"type":48,"value":145},{"type":43,"tag":147,"props":184,"children":186},{"className":185},[],[187],{"type":48,"value":188},"together-fine-tuning",{"type":48,"value":190}," for managed training jobs instead of raw cluster operations",{"type":43,"tag":51,"props":192,"children":194},{"id":193},"quick-routing",[195],{"type":48,"value":196},"Quick Routing",{"type":43,"tag":69,"props":198,"children":199},{},[200,240,268,288,307],{"type":43,"tag":73,"props":201,"children":202},{},[203,209],{"type":43,"tag":204,"props":205,"children":206},"strong",{},[207],{"type":48,"value":208},"Cluster creation, scaling, credentials, deletion",{"type":43,"tag":69,"props":210,"children":211},{},[212,230],{"type":43,"tag":73,"props":213,"children":214},{},[215,217,223,225],{"type":48,"value":216},"Start with ",{"type":43,"tag":218,"props":219,"children":221},"a",{"href":220},"scripts\u002Fmanage_cluster.py",[222],{"type":48,"value":220},{"type":48,"value":224}," or ",{"type":43,"tag":218,"props":226,"children":228},{"href":227},"scripts\u002Fmanage_cluster.ts",[229],{"type":48,"value":227},{"type":43,"tag":73,"props":231,"children":232},{},[233,235],{"type":48,"value":234},"Read ",{"type":43,"tag":218,"props":236,"children":238},{"href":237},"references\u002Fapi-reference.md",[239],{"type":48,"value":237},{"type":43,"tag":73,"props":241,"children":242},{},[243,248],{"type":43,"tag":204,"props":244,"children":245},{},[246],{"type":48,"value":247},"Shared storage lifecycle",{"type":43,"tag":69,"props":249,"children":250},{},[251,260],{"type":43,"tag":73,"props":252,"children":253},{},[254,255],{"type":48,"value":145},{"type":43,"tag":218,"props":256,"children":258},{"href":257},"scripts\u002Fmanage_storage.py",[259],{"type":48,"value":257},{"type":43,"tag":73,"props":261,"children":262},{},[263,264],{"type":48,"value":234},{"type":43,"tag":218,"props":265,"children":266},{"href":237},[267],{"type":48,"value":237},{"type":43,"tag":73,"props":269,"children":270},{},[271,276],{"type":43,"tag":204,"props":272,"children":273},{},[274],{"type":48,"value":275},"Kubernetes vs Slurm operations",{"type":43,"tag":69,"props":277,"children":278},{},[279],{"type":43,"tag":73,"props":280,"children":281},{},[282,283],{"type":48,"value":234},{"type":43,"tag":218,"props":284,"children":286},{"href":285},"references\u002Fcluster-management.md",[287],{"type":48,"value":285},{"type":43,"tag":73,"props":289,"children":290},{},[291,296],{"type":43,"tag":204,"props":292,"children":293},{},[294],{"type":48,"value":295},"Troubleshooting node health, PVCs, or scheduling",{"type":43,"tag":69,"props":297,"children":298},{},[299],{"type":43,"tag":73,"props":300,"children":301},{},[302,303],{"type":48,"value":234},{"type":43,"tag":218,"props":304,"children":305},{"href":285},[306],{"type":48,"value":285},{"type":43,"tag":73,"props":308,"children":309},{},[310,315],{"type":43,"tag":204,"props":311,"children":312},{},[313],{"type":48,"value":314},"Together CLI workflows",{"type":43,"tag":69,"props":316,"children":317},{},[318],{"type":43,"tag":73,"props":319,"children":320},{},[321,322],{"type":48,"value":234},{"type":43,"tag":218,"props":323,"children":325},{"href":324},"references\u002Fcli.md",[326],{"type":48,"value":324},{"type":43,"tag":51,"props":328,"children":330},{"id":329},"workflow",[331],{"type":48,"value":332},"Workflow",{"type":43,"tag":334,"props":335,"children":336},"ol",{},[337,342,347,352,357],{"type":43,"tag":73,"props":338,"children":339},{},[340],{"type":48,"value":341},"Decide whether the workload really needs cluster-level control.",{"type":43,"tag":73,"props":343,"children":344},{},[345],{"type":48,"value":346},"Choose on-demand vs reserved billing based on run duration and baseline utilization.",{"type":43,"tag":73,"props":348,"children":349},{},[350],{"type":48,"value":351},"Choose Kubernetes vs Slurm based on orchestration requirements and team tooling.",{"type":43,"tag":73,"props":353,"children":354},{},[355],{"type":48,"value":356},"Select region, GPU type, driver version, and shared storage plan.",{"type":43,"tag":73,"props":358,"children":359},{},[360],{"type":48,"value":361},"Provision first, then layer in access credentials, workload deployment, scaling, and health checks.",{"type":43,"tag":51,"props":363,"children":365},{"id":364},"high-signal-rules",[366],{"type":48,"value":367},"High-Signal Rules",{"type":43,"tag":69,"props":369,"children":370},{},[371,392,397,402,423,436,473,478,483,488],{"type":43,"tag":73,"props":372,"children":373},{},[374,376,382,384,390],{"type":48,"value":375},"Python scripts require the Together v2 SDK (",{"type":43,"tag":147,"props":377,"children":379},{"className":378},[],[380],{"type":48,"value":381},"together>=2.0.0",{"type":48,"value":383},"). If the user is on an older version, they must upgrade first: ",{"type":43,"tag":147,"props":385,"children":387},{"className":386},[],[388],{"type":48,"value":389},"uv pip install --upgrade \"together>=2.0.0\"",{"type":48,"value":391},".",{"type":43,"tag":73,"props":393,"children":394},{},[395],{"type":48,"value":396},"Prefer managed products unless the user explicitly needs raw infrastructure control.",{"type":43,"tag":73,"props":398,"children":399},{},[400],{"type":48,"value":401},"Treat storage lifecycle separately from cluster lifecycle; volumes can outlive clusters.",{"type":43,"tag":73,"props":403,"children":404},{},[405,407,413,415,421],{"type":48,"value":406},"When creating a cluster with new shared storage, prefer inline ",{"type":43,"tag":147,"props":408,"children":410},{"className":409},[],[411],{"type":48,"value":412},"shared_volume",{"type":48,"value":414}," over creating a volume separately and attaching via ",{"type":43,"tag":147,"props":416,"children":418},{"className":417},[],[419],{"type":48,"value":420},"volume_id",{"type":48,"value":422},". Separately created volumes may land in a different datacenter partition than the cluster, causing a \"does not exist in the datacenter\" error even when the volume shows as available.",{"type":43,"tag":73,"props":424,"children":425},{},[426,428,434],{"type":48,"value":427},"GPU stock-outs (409 \"Out of stock\") are common. Always call ",{"type":43,"tag":147,"props":429,"children":431},{"className":430},[],[432],{"type":48,"value":433},"list_regions()",{"type":48,"value":435}," first and be prepared to try multiple regions.",{"type":43,"tag":73,"props":437,"children":438},{},[439,441,447,449,455,457,463,465,471],{"type":48,"value":440},"The API requires ",{"type":43,"tag":147,"props":442,"children":444},{"className":443},[],[445],{"type":48,"value":446},"cuda_version",{"type":48,"value":448}," and ",{"type":43,"tag":147,"props":450,"children":452},{"className":451},[],[453],{"type":48,"value":454},"nvidia_driver_version",{"type":48,"value":456}," as separate fields in addition to the combined ",{"type":43,"tag":147,"props":458,"children":460},{"className":459},[],[461],{"type":48,"value":462},"driver_version",{"type":48,"value":464}," string. Pass them via ",{"type":43,"tag":147,"props":466,"children":468},{"className":467},[],[469],{"type":48,"value":470},"extra_body",{"type":48,"value":472}," in the Python SDK.",{"type":43,"tag":73,"props":474,"children":475},{},[476],{"type":48,"value":477},"Credentials retrieval is part of provisioning. Do not stop at cluster creation if the user needs to run workloads immediately.",{"type":43,"tag":73,"props":479,"children":480},{},[481],{"type":48,"value":482},"Slurm and Kubernetes operational patterns differ materially; read the cluster-management reference before improvising.",{"type":43,"tag":73,"props":484,"children":485},{},[486],{"type":48,"value":487},"For repeated cluster operations, start from the scripts instead of rebuilding request shapes.",{"type":43,"tag":73,"props":489,"children":490},{},[491,493,499,501,506,508,514,516,522,523,529],{"type":48,"value":492},"Slurm startup scripts (worker\u002Flogin init, worker\u002Fcontroller prolog and epilog, extra ",{"type":43,"tag":147,"props":494,"children":496},{"className":495},[],[497],{"type":48,"value":498},"slurm.conf",{"type":48,"value":500},") are ",{"type":43,"tag":204,"props":502,"children":503},{},[504],{"type":48,"value":505},"Slinky v1.0 only",{"type":48,"value":507},". A non-zero exit from a worker prolog or epilog drains the node, and calling Slurm commands (",{"type":43,"tag":147,"props":509,"children":511},{"className":510},[],[512],{"type":48,"value":513},"squeue",{"type":48,"value":515},", ",{"type":43,"tag":147,"props":517,"children":519},{"className":518},[],[520],{"type":48,"value":521},"scontrol",{"type":48,"value":515},{"type":43,"tag":147,"props":524,"children":526},{"className":525},[],[527],{"type":48,"value":528},"sacctmgr",{"type":48,"value":530},") inside any prolog\u002Fepilog can deadlock the scheduler.",{"type":43,"tag":51,"props":532,"children":534},{"id":533},"resource-map",[535],{"type":48,"value":536},"Resource Map",{"type":43,"tag":69,"props":538,"children":539},{},[540,554,567,580,593,606,619],{"type":43,"tag":73,"props":541,"children":542},{},[543,548,550],{"type":43,"tag":204,"props":544,"children":545},{},[546],{"type":48,"value":547},"Cluster API reference",{"type":48,"value":549},": ",{"type":43,"tag":218,"props":551,"children":552},{"href":237},[553],{"type":48,"value":237},{"type":43,"tag":73,"props":555,"children":556},{},[557,562,563],{"type":43,"tag":204,"props":558,"children":559},{},[560],{"type":48,"value":561},"Operational guide",{"type":48,"value":549},{"type":43,"tag":218,"props":564,"children":565},{"href":285},[566],{"type":48,"value":285},{"type":43,"tag":73,"props":568,"children":569},{},[570,575,576],{"type":43,"tag":204,"props":571,"children":572},{},[573],{"type":48,"value":574},"Operational troubleshooting",{"type":48,"value":549},{"type":43,"tag":218,"props":577,"children":578},{"href":285},[579],{"type":48,"value":285},{"type":43,"tag":73,"props":581,"children":582},{},[583,588,589],{"type":43,"tag":204,"props":584,"children":585},{},[586],{"type":48,"value":587},"CLI guide",{"type":48,"value":549},{"type":43,"tag":218,"props":590,"children":591},{"href":324},[592],{"type":48,"value":324},{"type":43,"tag":73,"props":594,"children":595},{},[596,601,602],{"type":43,"tag":204,"props":597,"children":598},{},[599],{"type":48,"value":600},"Python cluster management",{"type":48,"value":549},{"type":43,"tag":218,"props":603,"children":604},{"href":220},[605],{"type":48,"value":220},{"type":43,"tag":73,"props":607,"children":608},{},[609,614,615],{"type":43,"tag":204,"props":610,"children":611},{},[612],{"type":48,"value":613},"TypeScript cluster management",{"type":48,"value":549},{"type":43,"tag":218,"props":616,"children":617},{"href":227},[618],{"type":48,"value":227},{"type":43,"tag":73,"props":620,"children":621},{},[622,627,628],{"type":43,"tag":204,"props":623,"children":624},{},[625],{"type":48,"value":626},"Python storage management",{"type":48,"value":549},{"type":43,"tag":218,"props":629,"children":630},{"href":257},[631],{"type":48,"value":257},{"type":43,"tag":51,"props":633,"children":635},{"id":634},"official-docs",[636],{"type":48,"value":637},"Official Docs",{"type":43,"tag":69,"props":639,"children":640},{},[641,652,662,672,682],{"type":43,"tag":73,"props":642,"children":643},{},[644],{"type":43,"tag":218,"props":645,"children":649},{"href":646,"rel":647},"https:\u002F\u002Fdocs.together.ai\u002Fdocs\u002Fgpu-clusters-overview",[648],"nofollow",[650],{"type":48,"value":651},"GPU Clusters Overview",{"type":43,"tag":73,"props":653,"children":654},{},[655],{"type":43,"tag":218,"props":656,"children":659},{"href":657,"rel":658},"https:\u002F\u002Fdocs.together.ai\u002Fdocs\u002Fgpu-clusters-quickstart",[648],[660],{"type":48,"value":661},"GPU Clusters Quickstart",{"type":43,"tag":73,"props":663,"children":664},{},[665],{"type":43,"tag":218,"props":666,"children":669},{"href":667,"rel":668},"https:\u002F\u002Fdocs.together.ai\u002Freference\u002Fclusters-create",[648],[670],{"type":48,"value":671},"Clusters API",{"type":43,"tag":73,"props":673,"children":674},{},[675],{"type":43,"tag":218,"props":676,"children":679},{"href":677,"rel":678},"https:\u002F\u002Fdocs.together.ai\u002Fdocs\u002Fslurm-startup-scripts",[648],[680],{"type":48,"value":681},"Slurm Startup Scripts",{"type":43,"tag":73,"props":683,"children":684},{},[685],{"type":43,"tag":218,"props":686,"children":689},{"href":687,"rel":688},"https:\u002F\u002Fwww.together.ai\u002Finstant-gpu-clusters",[648],[690],{"type":48,"value":691},"Instant GPU Clusters",{"items":693,"total":863},[694,715,727,746,762,776,790,799,814,828,837,844],{"slug":695,"name":695,"fn":696,"description":697,"org":698,"tags":699,"stars":712,"repoUrl":713,"updatedAt":714},"blog-post","write structured blog posts","Write structured long-form blog posts with a consistent structure and SEO optimization.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[700,703,706,709],{"name":701,"slug":702,"type":16},"Content Creation","content-creation",{"name":704,"slug":705,"type":16},"Marketing","marketing",{"name":707,"slug":708,"type":16},"SEO","seo",{"name":710,"slug":711,"type":16},"Writing","writing",1151,"https:\u002F\u002Fgithub.com\u002Ftogethercomputer\u002Ftogether-cookbook","2026-07-17T06:07:39.475433",{"slug":716,"name":716,"fn":717,"description":718,"org":719,"tags":720,"stars":712,"repoUrl":713,"updatedAt":726},"social-media","create social media content","Create social media content including Twitter\u002FX threads, LinkedIn posts, and short-form updates.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[721,722,723,725],{"name":701,"slug":702,"type":16},{"name":704,"slug":705,"type":16},{"name":724,"slug":716,"type":16},"Social Media",{"name":710,"slug":711,"type":16},"2026-07-17T06:07:38.455058",{"slug":728,"name":728,"fn":729,"description":730,"org":731,"tags":732,"stars":26,"repoUrl":27,"updatedAt":745},"together-audio","process audio with Together AI","Text-to-speech and speech-to-text via Together AI, including REST, streaming, and realtime WebSocket TTS, plus transcription, translation, diarization, timestamps, and live STT. Reach for it whenever the user needs audio in or audio out on Together AI rather than chat generation, image or video creation, or model training.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[733,736,739,742],{"name":734,"slug":735,"type":16},"Audio","audio",{"name":737,"slug":738,"type":16},"Speech","speech",{"name":740,"slug":741,"type":16},"Text-to-Speech","text-to-speech",{"name":743,"slug":744,"type":16},"Transcription","transcription","2026-07-26T05:49:08.246858",{"slug":747,"name":747,"fn":748,"description":749,"org":750,"tags":751,"stars":26,"repoUrl":27,"updatedAt":761},"together-batch-inference","run asynchronous batch inference jobs","High-volume, asynchronous offline inference at up to 50% lower cost via Together AI's Batch API. Prepare JSONL inputs, upload files, create jobs, poll status, and download outputs. Reach for it whenever the user needs non-interactive bulk inference rather than real-time chat or evaluation jobs.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[752,755,758],{"name":753,"slug":754,"type":16},"Automation","automation",{"name":756,"slug":757,"type":16},"LLM","llm",{"name":759,"slug":760,"type":16},"Performance","performance","2026-07-17T06:08:23.919602",{"slug":763,"name":763,"fn":764,"description":765,"org":766,"tags":767,"stars":26,"repoUrl":27,"updatedAt":775},"together-chat-completions","generate text with Together AI","Real-time and streaming text generation via Together AI's OpenAI-compatible chat\u002Fcompletions API, including multi-turn conversations, tool and function calling, structured JSON outputs, and reasoning models. Reach for it whenever the user wants to build or debug text generation on Together AI, unless they specifically need batch jobs, embeddings, fine-tuning, dedicated endpoints, dedicated containers, or GPU clusters.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[768,771,774],{"name":769,"slug":770,"type":16},"AI","ai",{"name":772,"slug":773,"type":16},"API Development","api-development",{"name":756,"slug":757,"type":16},"2026-07-26T05:49:11.244104",{"slug":164,"name":164,"fn":777,"description":778,"org":779,"tags":780,"stars":26,"repoUrl":27,"updatedAt":789},"deploy custom inference containers on Together AI","Custom Dockerized inference workers on Together AI's managed GPU infrastructure. Build with Sprocket SDK, configure with Jig CLI, submit async queue jobs, and poll results. Reach for it whenever the user needs container-level control rather than a standard model endpoint or raw cluster.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[781,782,785,788],{"name":21,"slug":22,"type":16},{"name":783,"slug":784,"type":16},"Deployment","deployment",{"name":786,"slug":787,"type":16},"Docker","docker",{"name":14,"slug":15,"type":16},"2026-07-26T05:49:09.267892",{"slug":152,"name":152,"fn":791,"description":792,"org":793,"tags":794,"stars":26,"repoUrl":27,"updatedAt":798},"deploy and operate Together AI models","Deploy and operate models on dedicated GPUs with Together AI's Dedicated Model Inference (DMI, the v2 dedicated endpoints API): beta endpoints, deployments, deployment profiles and hardware configs, autoscaling, traffic splitting, A\u002FB tests, shadow experiments, Prometheus metrics, and custom model or LoRA adapter uploads. Reach for it whenever the user mentions together beta endpoints or tg beta commands, client.beta.endpoints, DMI resources like ep_\u002Fdep_\u002Fcr_\u002Fml_ IDs, or wants production model serving with traffic management on Together AI. This is the current dedicated-hosting API and also covers migrating off the retired legacy v1 endpoints API (non-beta client.endpoints \u002F together endpoints), whose create and restart now return HTTP 403.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[795,796,797],{"name":21,"slug":22,"type":16},{"name":783,"slug":784,"type":16},{"name":756,"slug":757,"type":16},"2026-07-26T06:08:44.044039",{"slug":800,"name":800,"fn":801,"description":802,"org":803,"tags":804,"stars":26,"repoUrl":27,"updatedAt":813},"together-embeddings","generate embeddings and build RAG pipelines","Dense vector embeddings, semantic search, RAG pipelines, and reranking via Together AI. Generate embeddings with open-source models and rerank results behind dedicated endpoints. Reach for it whenever the user needs vector representations or retrieval quality improvements rather than direct text generation.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[805,806,807,810],{"name":769,"slug":770,"type":16},{"name":756,"slug":757,"type":16},{"name":808,"slug":809,"type":16},"RAG","rag",{"name":811,"slug":812,"type":16},"Search","search","2026-07-26T05:49:06.247906",{"slug":815,"name":815,"fn":816,"description":817,"org":818,"tags":819,"stars":26,"repoUrl":27,"updatedAt":827},"together-evaluations","evaluate LLM outputs with Together AI","LLM-as-a-judge evaluation framework on Together AI. Classify, score, and compare model outputs, select judge models, use external-provider judges or targets, poll results and download reports. Reach for it whenever the user wants to benchmark outputs, grade responses, compare A\u002FB variants, or operationalize automated evaluations.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[820,823,826],{"name":821,"slug":822,"type":16},"Benchmarking","benchmarking",{"name":824,"slug":825,"type":16},"Evals","evals",{"name":756,"slug":757,"type":16},"2026-07-26T05:49:07.241553",{"slug":188,"name":188,"fn":829,"description":830,"org":831,"tags":832,"stars":26,"repoUrl":27,"updatedAt":836},"fine-tune and adapt models on Together AI","LoRA, full fine-tuning, DPO preference tuning, VLM training, function-calling tuning, reasoning tuning, and BYOM uploads on Together AI. Reach for it whenever the user wants to adapt a model on custom data rather than only run inference, evaluate outputs, or host an existing model.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[833,834,835],{"name":769,"slug":770,"type":16},{"name":756,"slug":757,"type":16},{"name":14,"slug":15,"type":16},"2026-07-26T05:49:10.243114",{"slug":4,"name":4,"fn":5,"description":6,"org":838,"tags":839,"stars":26,"repoUrl":27,"updatedAt":28},{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[840,841,842,843],{"name":21,"slug":22,"type":16},{"name":18,"slug":19,"type":16},{"name":24,"slug":25,"type":16},{"name":14,"slug":15,"type":16},{"slug":845,"name":845,"fn":846,"description":847,"org":848,"tags":849,"stars":26,"repoUrl":27,"updatedAt":862},"together-images","generate and edit images with Together AI","Text-to-image generation and image editing via Together AI, including FLUX and Kontext models, LoRA-based styling, reference-image guidance, and local image downloads. Reach for it whenever the user wants to generate or edit images on Together AI rather than create videos or build text-only chat applications.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[850,853,856,859],{"name":851,"slug":852,"type":16},"Creative","creative",{"name":854,"slug":855,"type":16},"Design","design",{"name":857,"slug":858,"type":16},"Image Generation","image-generation",{"name":860,"slug":861,"type":16},"Multimodal","multimodal","2026-07-17T06:04:23.067755",16,{"items":865,"total":911},[866,873,879,885,892,898,905],{"slug":728,"name":728,"fn":729,"description":730,"org":867,"tags":868,"stars":26,"repoUrl":27,"updatedAt":745},{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[869,870,871,872],{"name":734,"slug":735,"type":16},{"name":737,"slug":738,"type":16},{"name":740,"slug":741,"type":16},{"name":743,"slug":744,"type":16},{"slug":747,"name":747,"fn":748,"description":749,"org":874,"tags":875,"stars":26,"repoUrl":27,"updatedAt":761},{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[876,877,878],{"name":753,"slug":754,"type":16},{"name":756,"slug":757,"type":16},{"name":759,"slug":760,"type":16},{"slug":763,"name":763,"fn":764,"description":765,"org":880,"tags":881,"stars":26,"repoUrl":27,"updatedAt":775},{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[882,883,884],{"name":769,"slug":770,"type":16},{"name":772,"slug":773,"type":16},{"name":756,"slug":757,"type":16},{"slug":164,"name":164,"fn":777,"description":778,"org":886,"tags":887,"stars":26,"repoUrl":27,"updatedAt":789},{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[888,889,890,891],{"name":21,"slug":22,"type":16},{"name":783,"slug":784,"type":16},{"name":786,"slug":787,"type":16},{"name":14,"slug":15,"type":16},{"slug":152,"name":152,"fn":791,"description":792,"org":893,"tags":894,"stars":26,"repoUrl":27,"updatedAt":798},{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[895,896,897],{"name":21,"slug":22,"type":16},{"name":783,"slug":784,"type":16},{"name":756,"slug":757,"type":16},{"slug":800,"name":800,"fn":801,"description":802,"org":899,"tags":900,"stars":26,"repoUrl":27,"updatedAt":813},{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[901,902,903,904],{"name":769,"slug":770,"type":16},{"name":756,"slug":757,"type":16},{"name":808,"slug":809,"type":16},{"name":811,"slug":812,"type":16},{"slug":815,"name":815,"fn":816,"description":817,"org":906,"tags":907,"stars":26,"repoUrl":27,"updatedAt":827},{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[908,909,910],{"name":821,"slug":822,"type":16},{"name":824,"slug":825,"type":16},{"name":756,"slug":757,"type":16},14]