Proposal introducing new Qwen3 models

This proposal introduces new Qwen3 models including Qwen3-32B-FP8 and Qwen3-235B-A22B-Instruct-2507-FP8, along with updating parameters for Qwen2.5-7B-Instruct and QwQ-32B.

Proposal Messages (5)

[
  {
    "@type": "/inference.inference.MsgRegisterModel",
    "authority": "gonka10d07y265gmmuvt4z0w9aw880jnsr700j2h5m33",
    "proposed_by": "gonka10d07y265gmmuvt4z0w9aw880jnsr700j2h5m33",
    "id": "Qwen/Qwen2.5-7B-Instruct",
    "units_of_compute_per_token": "100",
    "hf_repo": "Qwen/Qwen2.5-7B-Instruct",
    "hf_commit": "a09a35458c702b33eeacc393d103063234e8bc28",
    "model_args": [
      "--quantization",
      "fp8"
    ],
    "v_ram": "24",
    "throughput_per_nonce": "20000",
    "validation_threshold": {
      "value": "97",
      "exponent": -2
    }
  },
  {
    "@type": "/inference.inference.MsgRegisterModel",
    "authority": "gonka10d07y265gmmuvt4z0w9aw880jnsr700j2h5m33",
    "proposed_by": "gonka10d07y265gmmuvt4z0w9aw880jnsr700j2h5m33",
    "id": "Qwen/QwQ-32B",
    "units_of_compute_per_token": "1000",
    "hf_repo": "Qwen/QwQ-32B",
    "hf_commit": "976055f8c83f394f35dbd3ab09a285a984907bd0",
    "model_args": [
      "--quantization",
      "fp8",
      "--kv-cache-dtype",
      "fp8"
    ],
    "v_ram": "80",
    "throughput_per_nonce": "6000",
    "validation_threshold": {
      "value": "97",
      "exponent": -2
    }
  },
  {
    "@type": "/inference.inference.MsgRegisterModel",
    "authority": "gonka10d07y265gmmuvt4z0w9aw880jnsr700j2h5m33",
    "proposed_by": "gonka1kx9mca3xm8u8ypzfuhmxey66u0ufxhs7nm6wc5",
    "id": "RedHatAI/Qwen2.5-7B-Instruct-quantized.w8a16",
    "units_of_compute_per_token": "100",
    "hf_repo": "RedHatAI/Qwen2.5-7B-Instruct-quantized.w8a16",
    "hf_commit": "a5609cbf939e46f4c17f189100d5e6825da318b0",
    "model_args": [],
    "v_ram": "24",
    "throughput_per_nonce": "20000",
    "validation_threshold": {
      "value": "991429",
      "exponent": -6
    }
  },
  {
    "@type": "/inference.inference.MsgRegisterModel",
    "authority": "gonka10d07y265gmmuvt4z0w9aw880jnsr700j2h5m33",
    "proposed_by": "gonka1kx9mca3xm8u8ypzfuhmxey66u0ufxhs7nm6wc5",
    "id": "Qwen/Qwen3-32B-FP8",
    "units_of_compute_per_token": "1000",
    "hf_repo": "Qwen/Qwen3-32B-FP8",
    "hf_commit": "aa55da1ecc13d006e8b8e4f54579b1ea8c3db2df",
    "model_args": [],
    "v_ram": "80",
    "throughput_per_nonce": "6000",
    "validation_threshold": {
      "value": "95814",
      "exponent": -5
    }
  },
  {
    "@type": "/inference.inference.MsgRegisterModel",
    "authority": "gonka10d07y265gmmuvt4z0w9aw880jnsr700j2h5m33",
    "proposed_by": "gonka1kx9mca3xm8u8ypzfuhmxey66u0ufxhs7nm6wc5",
    "id": "Qwen/Qwen3-235B-A22B-Instruct-2507-FP8",
    "units_of_compute_per_token": "10000",
    "hf_repo": "Qwen/Qwen3-235B-A22B-Instruct-2507-FP8",
    "hf_commit": "c0eb82898e3da8fb6dd017e3e6698a5e37b3a3e6",
    "model_args": [
      "--max-model-len",
      "240000"
    ],
    "v_ram": "320",
    "throughput_per_nonce": "1500",
    "validation_threshold": {
      "value": "970917",
      "exponent": -6
    }
  }
]