{
  "schema": "infervisor.models.v1",
  "generated": "2026-11-04T03:35:35.871Z",
  "site": "https://infervisor.ai",
  "registry": {
    "url": "https://pdist.infervisor.ai",
    "catalog": "https://pdist.infervisor.ai/catalog.json",
    "signing_key": "infervisor-2026:ESz+iXQToovy4GhRUgTHoUKUwMumUA/ls/4tAaEjL3M=",
    "install": "curl -fsSL https://pdist.infervisor.ai/install.sh | sh",
    "runtime": "0.1.3+d371228"
  },
  "endpoints": {
    "models": "https://infervisor.ai/models.json",
    "model": "https://infervisor.ai/models/{name}.json",
    "llms_txt": "https://infervisor.ai/llms.txt"
  },
  "models": [
    {
      "name": "gemma-4-31b",
      "display": "Gemma 4 31B",
      "task": "text",
      "task_label": "text generation",
      "summary": "The largest dense Gemma 4, instruction tuned, in bf16 and FP8 on NVIDIA.",
      "typical": "The strongest dense Gemma 4 for chat, summarisation and long prompts on one H100.",
      "params": "31B",
      "license": "Gemma Terms of Use",
      "gated": true,
      "voice": null,
      "hf": "google/gemma-4-31B-it",
      "hf_url": "https://huggingface.co/google/gemma-4-31B-it",
      "revision": "",
      "checkpoints": [
        {
          "hf": "RedHatAI/gemma-4-31B-it-FP8-dynamic",
          "hf_url": "https://huggingface.co/RedHatAI/gemma-4-31B-it-FP8-dynamic/tree/d4ab4f579dd3516f97d8a6a4c98d0653480bad15",
          "revision": "d4ab4f579dd3516f97d8a6a4c98d0653480bad15",
          "precision": [
            "fp8"
          ],
          "builds": [
            "sm90a-h100-tp1-fp8"
          ]
        }
      ],
      "page": "https://infervisor.ai/models/gemma-4-31b",
      "json": "https://infervisor.ai/models/gemma-4-31b.json",
      "commands": {
        "install": "curl -fsSL https://pdist.infervisor.ai/install.sh | sh",
        "load": "export HF_TOKEN=<your Hugging Face token>   # gated checkpoint\nplowrt load gemma-4-31b --fetch-weights",
        "serve": "plowrt serve --model gemma-4-31b --port 8080",
        "request": "curl -s http://127.0.0.1:8080/v1/chat/completions \\\n  -H 'content-type: application/json' \\\n  -d '{\"model\":\"gemma-4-31b\",\"messages\":[{\"role\":\"user\",\"content\":\"Hello\"}]}'"
      },
      "max_ctx": 262144,
      "context_source": "https://huggingface.co/google/gemma-4-31B-it/blob/main/config.json",
      "builds": [
        {
          "label": "sm90a-h100-tp1-fp8",
          "generation": 1,
          "status": "emits",
          "hardware": "NVIDIA H100 SXM5",
          "vendor": "nvidia",
          "isa": "sm_90a",
          "sku": "H100 SXM5",
          "units": 132,
          "mem_bytes": 85899345920,
          "also_runs_on": [
            "H100 NVL",
            "H200 SXM"
          ],
          "parallel": {
            "mode": "tp",
            "n": 1
          },
          "max_ctx": 16384,
          "precision": [
            "fp8"
          ],
          "bytes": 185791348,
          "released": "2026-10-29",
          "checkpoint": {
            "hf": "RedHatAI/gemma-4-31B-it-FP8-dynamic",
            "revision": "d4ab4f579dd3516f97d8a6a4c98d0653480bad15"
          },
          "load": "plowrt load infervisor/gemma-4-31b:sm90a-h100-tp1-fp8",
          "pinned": "plowrt load infervisor/gemma-4-31b:sm90a-h100-tp1-fp8@g1",
          "manifest": "https://pdist.infervisor.ai/v1/infervisor/gemma-4-31b/manifests/sm90a-h100-tp1-fp8@g1"
        }
      ]
    },
    {
      "name": "gemma-4-e4b",
      "display": "Gemma 4 E4B",
      "task": "text",
      "task_label": "text generation",
      "summary": "Google's small instruction-tuned Gemma 4, the text model of the voice-agent stack.",
      "typical": "The text model of a voice agent: short multi-turn prompts, many concurrent sessions, logprobs on.",
      "params": "4B effective (8B with per-layer embeddings)",
      "license": "Gemma Terms of Use",
      "gated": true,
      "voice": null,
      "hf": "google/gemma-4-E4B-it",
      "hf_url": "https://huggingface.co/google/gemma-4-E4B-it/tree/ee0ef6023621cff504d758262d4e04895a5af4a2",
      "revision": "ee0ef6023621cff504d758262d4e04895a5af4a2",
      "checkpoints": [],
      "page": "https://infervisor.ai/models/gemma-4-e4b",
      "json": "https://infervisor.ai/models/gemma-4-e4b.json",
      "commands": {
        "install": "curl -fsSL https://pdist.infervisor.ai/install.sh | sh",
        "load": "export HF_TOKEN=<your Hugging Face token>   # gated checkpoint\nplowrt load gemma-4-e4b --fetch-weights",
        "serve": "plowrt serve --model gemma-4-e4b --port 8080",
        "request": "curl -s http://127.0.0.1:8080/v1/chat/completions \\\n  -H 'content-type: application/json' \\\n  -d '{\"model\":\"gemma-4-e4b\",\"messages\":[{\"role\":\"user\",\"content\":\"Hello\"}]}'"
      },
      "max_ctx": 131072,
      "context_source": "https://huggingface.co/google/gemma-4-E4B-it/blob/main/config.json",
      "builds": [
        {
          "label": "sm90a-h100-tp1",
          "generation": 1,
          "status": "emits",
          "hardware": "NVIDIA H100 SXM5",
          "vendor": "nvidia",
          "isa": "sm_90a",
          "sku": "H100 SXM5",
          "units": 132,
          "mem_bytes": 85899345920,
          "also_runs_on": [
            "H100 NVL",
            "H200 SXM"
          ],
          "parallel": {
            "mode": "tp",
            "n": 1
          },
          "max_ctx": 8192,
          "precision": [],
          "bytes": 134271834,
          "released": "2026-10-29",
          "checkpoint": {
            "hf": "google/gemma-4-E4B-it",
            "revision": "ee0ef6023621cff504d758262d4e04895a5af4a2"
          },
          "load": "plowrt load infervisor/gemma-4-e4b:sm90a-h100-tp1",
          "pinned": "plowrt load infervisor/gemma-4-e4b:sm90a-h100-tp1@g1",
          "manifest": "https://pdist.infervisor.ai/v1/infervisor/gemma-4-e4b/manifests/sm90a-h100-tp1@g1"
        }
      ]
    },
    {
      "name": "orpheus",
      "display": "Orpheus 3B",
      "task": "tts",
      "task_label": "text to speech",
      "summary": "Canopy Labs' Orpheus 3B, a Llama-based codec LM with a SNAC decoder, served with the same codec-LM contract as Veena.",
      "typical": "Expressive English speech synthesis from a 3B codec LM.",
      "params": "3B",
      "license": "Apache-2.0",
      "gated": false,
      "voice": "tara",
      "hf": "unsloth/orpheus-3b-0.1-ft",
      "hf_url": "https://huggingface.co/unsloth/orpheus-3b-0.1-ft/tree/eae2b6e5e429c81b95ac42a883ac64f126583d43",
      "revision": "eae2b6e5e429c81b95ac42a883ac64f126583d43",
      "checkpoints": [],
      "page": "https://infervisor.ai/models/orpheus",
      "json": "https://infervisor.ai/models/orpheus.json",
      "commands": {
        "install": "curl -fsSL https://pdist.infervisor.ai/install.sh | sh",
        "load": "plowrt load orpheus --fetch-weights",
        "serve": "plowrt serve --model orpheus --port 8080",
        "request": "curl -s http://127.0.0.1:8080/v1/audio/speech \\\n  -H 'content-type: application/json' \\\n  -d '{\"model\":\"orpheus\",\"input\":\"Hello from plow.\",\"voice\":\"tara\"}' -o out.wav"
      },
      "max_ctx": 131072,
      "context_source": "https://huggingface.co/unsloth/orpheus-3b-0.1-ft/blob/main/config.json",
      "builds": [
        {
          "label": "sm90a-h100-tp1",
          "generation": 1,
          "status": "emits",
          "hardware": "NVIDIA H100 SXM5",
          "vendor": "nvidia",
          "isa": "sm_90a",
          "sku": "H100 SXM5",
          "units": 132,
          "mem_bytes": 85899345920,
          "also_runs_on": [
            "H100 NVL",
            "H200 SXM"
          ],
          "parallel": {
            "mode": "tp",
            "n": 1
          },
          "max_ctx": 2048,
          "precision": [],
          "bytes": 197439782,
          "released": "2026-10-29",
          "checkpoint": {
            "hf": "unsloth/orpheus-3b-0.1-ft",
            "revision": "eae2b6e5e429c81b95ac42a883ac64f126583d43"
          },
          "load": "plowrt load infervisor/orpheus:sm90a-h100-tp1",
          "pinned": "plowrt load infervisor/orpheus:sm90a-h100-tp1@g1",
          "manifest": "https://pdist.infervisor.ai/v1/infervisor/orpheus/manifests/sm90a-h100-tp1@g1"
        }
      ]
    },
    {
      "name": "qwen3.8-27b",
      "display": "Qwen3.8 27B",
      "task": "text",
      "task_label": "text generation",
      "summary": "Qwen3.8-27B with gated DeltaNet recurrent layers, in bf16 and FP8 on NVIDIA.",
      "typical": "Chat and agent workloads on one H100 with a long decode ladder.",
      "params": "27B",
      "license": "Apache-2.0",
      "gated": false,
      "voice": null,
      "hf": "Qwen/Qwen3.8-27B",
      "hf_url": "https://huggingface.co/Qwen/Qwen3.8-27B/tree/1d4bf0f2ff6012fd82039f2fa52739d0dd7c60c0",
      "revision": "1d4bf0f2ff6012fd82039f2fa52739d0dd7c60c0",
      "checkpoints": [],
      "page": "https://infervisor.ai/models/qwen3.8-27b",
      "json": "https://infervisor.ai/models/qwen3.8-27b.json",
      "commands": {
        "install": "curl -fsSL https://pdist.infervisor.ai/install.sh | sh",
        "load": "plowrt load qwen3.8-27b --fetch-weights",
        "serve": "plowrt serve --model qwen3.8-27b --port 8080",
        "request": "curl -s http://127.0.0.1:8080/v1/chat/completions \\\n  -H 'content-type: application/json' \\\n  -d '{\"model\":\"qwen3.8-27b\",\"messages\":[{\"role\":\"user\",\"content\":\"Hello\"}]}'"
      },
      "max_ctx": 262144,
      "context_source": "https://huggingface.co/Qwen/Qwen3.8-27B/blob/main/config.json",
      "builds": [
        {
          "label": "sm90a-h100-tp1",
          "generation": 1,
          "status": "emits",
          "hardware": "NVIDIA H100 SXM5",
          "vendor": "nvidia",
          "isa": "sm_90a",
          "sku": "H100 SXM5",
          "units": 132,
          "mem_bytes": 85899345920,
          "also_runs_on": [
            "H100 NVL",
            "H200 SXM"
          ],
          "parallel": {
            "mode": "tp",
            "n": 1
          },
          "max_ctx": 16384,
          "precision": [],
          "bytes": 97532832,
          "released": "2026-10-29",
          "checkpoint": {
            "hf": "Qwen/Qwen3.8-27B",
            "revision": "1d4bf0f2ff6012fd82039f2fa52739d0dd7c60c0"
          },
          "load": "plowrt load infervisor/qwen3.8-27b:sm90a-h100-tp1",
          "pinned": "plowrt load infervisor/qwen3.8-27b:sm90a-h100-tp1@g1",
          "manifest": "https://pdist.infervisor.ai/v1/infervisor/qwen3.8-27b/manifests/sm90a-h100-tp1@g1"
        }
      ]
    },
    {
      "name": "qwen3-0.6b",
      "display": "Qwen3 0.6B",
      "task": "text",
      "task_label": "text generation",
      "summary": "Qwen's smallest Qwen3, published for Apple Silicon.",
      "typical": "A tiny chat model for an Apple Silicon laptop or desk machine.",
      "params": "0.6B",
      "license": "Apache-2.0",
      "gated": false,
      "voice": null,
      "hf": "Qwen/Qwen3-0.6B",
      "hf_url": "https://huggingface.co/Qwen/Qwen3-0.6B/tree/c1899de289a04d12100db370d81485cdf75e47ca",
      "revision": "c1899de289a04d12100db370d81485cdf75e47ca",
      "checkpoints": [],
      "page": "https://infervisor.ai/models/qwen3-0.6b",
      "json": "https://infervisor.ai/models/qwen3-0.6b.json",
      "commands": {
        "install": "curl -fsSL https://pdist.infervisor.ai/install.sh | sh",
        "load": "plowrt load qwen3-0.6b --fetch-weights",
        "serve": "plowrt serve --model qwen3-0.6b --port 8080",
        "request": "curl -s http://127.0.0.1:8080/v1/chat/completions \\\n  -H 'content-type: application/json' \\\n  -d '{\"model\":\"qwen3-0.6b\",\"messages\":[{\"role\":\"user\",\"content\":\"Hello\"}]}'"
      },
      "max_ctx": 32768,
      "context_source": "https://huggingface.co/Qwen/Qwen3-0.6B",
      "builds": [
        {
          "label": "metal3-apple-m4-pro-tp1-2k",
          "generation": 1,
          "status": "emits",
          "hardware": "Apple M4 Pro",
          "vendor": "apple",
          "isa": "metal3",
          "sku": "Apple M4 Pro",
          "units": 16,
          "mem_bytes": 19069665280,
          "also_runs_on": [
            "Apple M4 Max"
          ],
          "parallel": {
            "mode": "tp",
            "n": 1
          },
          "max_ctx": 2048,
          "precision": [],
          "bytes": 14862702,
          "released": "2026-10-28",
          "checkpoint": {
            "hf": "Qwen/Qwen3-0.6B",
            "revision": "c1899de289a04d12100db370d81485cdf75e47ca"
          },
          "load": "plowrt load infervisor/qwen3-0.6b:metal3-apple-m4-pro-tp1-2k",
          "pinned": "plowrt load infervisor/qwen3-0.6b:metal3-apple-m4-pro-tp1-2k@g1",
          "manifest": "https://pdist.infervisor.ai/v1/infervisor/qwen3-0.6b/manifests/metal3-apple-m4-pro-tp1-2k@g1"
        }
      ]
    },
    {
      "name": "qwen3-asr-0.6b",
      "display": "Qwen3 ASR 0.6B",
      "task": "asr",
      "task_label": "speech to text",
      "summary": "The 0.6B Qwen3 ASR model, for latency-first transcription on a small GPU.",
      "typical": "Latency-first transcription on a single small GPU.",
      "params": "0.6B",
      "license": "Apache-2.0",
      "gated": false,
      "voice": null,
      "hf": "Qwen/Qwen3-ASR-0.6B",
      "hf_url": "https://huggingface.co/Qwen/Qwen3-ASR-0.6B/tree/5eb144179a02acc5e5ba31e748d22b0cf3e303b0",
      "revision": "5eb144179a02acc5e5ba31e748d22b0cf3e303b0",
      "checkpoints": [],
      "page": "https://infervisor.ai/models/qwen3-asr-0.6b",
      "json": "https://infervisor.ai/models/qwen3-asr-0.6b.json",
      "commands": {
        "install": "curl -fsSL https://pdist.infervisor.ai/install.sh | sh",
        "load": "plowrt load qwen3-asr-0.6b --fetch-weights",
        "serve": "plowrt serve --model qwen3-asr-0.6b --port 8080",
        "request": "curl -s http://127.0.0.1:8080/v1/audio/transcriptions \\\n  -F model=qwen3-asr-0.6b -F file=@sample.wav"
      },
      "max_ctx": 65536,
      "context_source": "https://huggingface.co/Qwen/Qwen3-ASR-0.6B/blob/main/config.json",
      "builds": [
        {
          "label": "sm90a-h100-tp1",
          "generation": 1,
          "status": "emits",
          "hardware": "NVIDIA H100 SXM5",
          "vendor": "nvidia",
          "isa": "sm_90a",
          "sku": "H100 SXM5",
          "units": 132,
          "mem_bytes": 85899345920,
          "also_runs_on": [
            "H100 NVL",
            "H200 SXM"
          ],
          "parallel": {
            "mode": "tp",
            "n": 1
          },
          "max_ctx": 2048,
          "precision": [],
          "bytes": 477310090,
          "released": "2026-10-28",
          "checkpoint": {
            "hf": "Qwen/Qwen3-ASR-0.6B",
            "revision": "5eb144179a02acc5e5ba31e748d22b0cf3e303b0"
          },
          "load": "plowrt load infervisor/qwen3-asr-0.6b:sm90a-h100-tp1",
          "pinned": "plowrt load infervisor/qwen3-asr-0.6b:sm90a-h100-tp1@g1",
          "manifest": "https://pdist.infervisor.ai/v1/infervisor/qwen3-asr-0.6b/manifests/sm90a-h100-tp1@g1"
        }
      ]
    },
    {
      "name": "veena",
      "display": "Veena",
      "task": "tts",
      "task_label": "text to speech",
      "summary": "Maya Research's Veena, a Llama-style codec language model with a SNAC decoder for Hindi and English speech.",
      "typical": "Hindi and English speech synthesis with a Llama-style codec LM.",
      "params": "3B",
      "license": "Apache-2.0",
      "gated": false,
      "voice": "kavya",
      "hf": "maya-research/Veena",
      "hf_url": "https://huggingface.co/maya-research/Veena/tree/8b770f9e69e6b35ef320d4cd70a99a4ab6dd022f",
      "revision": "8b770f9e69e6b35ef320d4cd70a99a4ab6dd022f",
      "checkpoints": [],
      "page": "https://infervisor.ai/models/veena",
      "json": "https://infervisor.ai/models/veena.json",
      "commands": {
        "install": "curl -fsSL https://pdist.infervisor.ai/install.sh | sh",
        "load": "plowrt load veena --fetch-weights",
        "serve": "plowrt serve --model veena --port 8080",
        "request": "curl -s http://127.0.0.1:8080/v1/audio/speech \\\n  -H 'content-type: application/json' \\\n  -d '{\"model\":\"veena\",\"input\":\"Hello from plow.\",\"voice\":\"kavya\"}' -o out.wav"
      },
      "max_ctx": 131072,
      "context_source": "https://huggingface.co/maya-research/Veena/blob/main/config.json",
      "builds": [
        {
          "label": "sm90a-h100-tp1",
          "generation": 1,
          "status": "emits",
          "hardware": "NVIDIA H100 SXM5",
          "vendor": "nvidia",
          "isa": "sm_90a",
          "sku": "H100 SXM5",
          "units": 132,
          "mem_bytes": 85899345920,
          "also_runs_on": [
            "H100 NVL",
            "H200 SXM"
          ],
          "parallel": {
            "mode": "tp",
            "n": 1
          },
          "max_ctx": 2048,
          "precision": [],
          "bytes": 146029210,
          "released": "2026-10-28",
          "checkpoint": {
            "hf": "maya-research/Veena",
            "revision": "8b770f9e69e6b35ef320d4cd70a99a4ab6dd022f"
          },
          "load": "plowrt load infervisor/veena:sm90a-h100-tp1",
          "pinned": "plowrt load infervisor/veena:sm90a-h100-tp1@g1",
          "manifest": "https://pdist.infervisor.ai/v1/infervisor/veena/manifests/sm90a-h100-tp1@g1"
        }
      ]
    }
  ]
}