{
  "schema": 1,
  "version": 4,
  "published": "2026-10-29",
  "models": [
    {
      "id": "qwen3.6-35b-a3b-q3km",
      "name": "Qwen3.6 35B-A3B (Q3_K_M)",
      "family": "qwen3.6",
      "description": "Qwen's 35B mixture-of-experts reasoning model (3B active) on one Tesla K80 board, with MTP speculative decoding.",
      "licence": {
        "id": "apache-2.0",
        "url": "https://huggingface.co/Qwen/Qwen3.6-35B-A3B/blob/main/LICENSE"
      },
      "capabilities": [
        "chat",
        "tools",
        "reasoning"
      ],
      "files": [
        {
          "role": "weights",
          "hf_repo": "TeamIDE/Qwen3.6-35B-A3B-Pantheon-K80",
          "revision": "b4f4e2c4ee73c1a4b2c2149ffa4912491d1ac517",
          "filename": "Qwen_Qwen3.6-35B-A3B-Q3_K_M.gguf",
          "size": 17117951616,
          "sha256": "17350b1355d32397cffe7d4b446fd9361dbb01aa26bbf41fc950aa595033794f"
        }
      ],
      "profiles": [
        {
          "id": "k80-pair",
          "rank": 1,
          "hardware": {
            "unit_type": "k80-gpu",
            "count": 2,
            "same_card": true,
            "min_mem_per_unit": 10756292608,
            "min_host_ram": 7000000000
          },
          "load": {
            "ctx_cap": 40960,
            "max_sessions": 1,
            "max_concurrent": 0,
            "reasoning_budget": 2000,
            "mtp": true
          },
          "options": {
            "kv_type": "q4",
            "ctx_min": 8192,
            "ctx_max": 131072,
            "sessions_max": 4,
            "mem": {
              "fixed": 12075220992,
              "ctx_per_token": 3478,
              "per_session": 4718592,
              "kv_per_token": 2880
            }
          },
          "sampling": {
            "temperature": 0.6,
            "top_p": 0.95,
            "top_k": 20,
            "min_p": 0,
            "presence_penalty": 0
          },
          "measured": {
            "decode_tok_s": {
              "8k": 24.5,
              "14k": 19.7,
              "32k": 13.6
            },
            "ttft": {
              "prompt_tokens": 8444,
              "seconds": 74.2
            },
            "context_window": 40960
          }
        },
        {
          "id": "k80-pair-fp32",
          "rank": 2,
          "hardware": {
            "unit_type": "k80-gpu",
            "count": 2,
            "same_card": true,
            "min_mem_per_unit": 10756292608,
            "min_host_ram": 7000000000
          },
          "load": {
            "ctx_cap": 32768,
            "max_sessions": 1,
            "max_concurrent": 0,
            "reasoning_budget": 2000,
            "mtp": false
          },
          "options": {
            "kv_type": "fp32",
            "ctx_min": 8192,
            "ctx_max": 32768,
            "sessions_max": 4,
            "mem": {
              "fixed": 10800332800,
              "ctx_per_token": 20480,
              "per_session": 36700160,
              "kv_per_token": 20480
            }
          },
          "sampling": {
            "temperature": 0.6,
            "top_p": 0.95,
            "top_k": 20,
            "min_p": 0,
            "presence_penalty": 0
          },
          "measured": {
            "decode_tok_s": {
              "8k": 19.13,
              "14k": 16.16,
              "32k": 0
            },
            "ttft": {
              "prompt_tokens": 8444,
              "seconds": 102.0
            },
            "context_window": 32768
          }
        }
      ]
    },
    {
      "id": "qwen3.6-35b-a3b-q8",
      "name": "Qwen3.6 35B-A3B (Q8_0)",
      "family": "qwen3.6",
      "description": "Qwen's 35B mixture-of-experts reasoning model (3B active) from Q8_0 weights, on one Tenstorrent Blackhole p150 or two Tesla K80 boards (4 GPUs), with MTP speculative decoding.",
      "licence": {
        "id": "apache-2.0",
        "url": "https://huggingface.co/Qwen/Qwen3.6-35B-A3B/blob/main/LICENSE"
      },
      "capabilities": [
        "chat",
        "tools",
        "reasoning"
      ],
      "files": [
        {
          "role": "weights",
          "hf_repo": "TeamIDE/Qwen3.6-35B-A3B-Pantheon-K80",
          "revision": "272bf5edbc1255971effa8d45e62e13b9398ba54",
          "filename": "Qwen3.6-35B-A3B-Q8_0.gguf",
          "size": 37801097248,
          "sha256": "0bd4666292101b41b1965008814f0948c1a6a7e0bd7187e7a12ce6dd1d16212e",
          "unit_types": ["k80-gpu"]
        },
        {
          "role": "weights",
          "hf_repo": "TeamIDE/Qwen3.6-35B-A3B-Pantheon-TT",
          "revision": "3cf0ba295349218c2ac28f82ba25f3142260eb97",
          "filename": "Qwen_Qwen3.6-35B-A3B-Q8_0.meta.gguf",
          "size": 15972512,
          "sha256": "90f46deabc2feda8ed32af9f83c96d1709bc1148040142b48bd2b4b09830b01c",
          "unit_types": ["p150"]
        },
        {
          "role": "extra",
          "hf_repo": "TeamIDE/Qwen3.6-35B-A3B-Pantheon-TT",
          "revision": "3cf0ba295349218c2ac28f82ba25f3142260eb97",
          "filename": "Qwen_Qwen3.6-35B-A3B-Q8_0.meta.moe_bfp6.fttw",
          "size": 28843656936,
          "sha256": "deec8f87dcb7b664aac4537aa00b9fa6a1094deae3af4dbb2a3ecd7deea137bf",
          "unit_types": ["p150"]
        },
        {
          "role": "extra",
          "hf_repo": "TeamIDE/Qwen3.6-35B-A3B-Pantheon-TT",
          "revision": "3cf0ba295349218c2ac28f82ba25f3142260eb97",
          "filename": "Qwen_Qwen3.6-35B-A3B-Q8_0.meta.moe_bfp6.mtp.fttw",
          "size": 697686184,
          "sha256": "cd394ede33894a8ca4840f9fa57fc9d4c165124776e9983065072c32e9a659ab",
          "unit_types": ["p150"]
        }
      ],
      "profiles": [
        {
          "id": "p150-one",
          "rank": 1,
          "hardware": {
            "unit_type": "p150",
            "count": 1,
            "same_card": false,
            "min_mem_per_unit": 0,
            "min_host_ram": 7000000000
          },
          "load": {
            "ctx_cap": 36864,
            "max_sessions": 1,
            "max_concurrent": 0,
            "reasoning_budget": 0,
            "mtp": true
          },
          "options": {
            "kv_type": "bf16",
            "ctx_min": 8192,
            "ctx_max": 36864,
            "sessions_max": 1,
            "mem": {
              "fixed": 31370000000,
              "ctx_per_token": 0,
              "per_session": 0,
              "kv_per_token": 22528
            }
          },
          "sampling": {
            "temperature": 0.6,
            "top_p": 0.95,
            "top_k": 20,
            "min_p": 0,
            "presence_penalty": 0
          },
          "measured": {
            "decode_tok_s": {
              "8k": 12.18,
              "14k": 11.43,
              "32k": 9.83
            },
            "ttft": {
              "prompt_tokens": 8147,
              "seconds": 192.0
            },
            "context_window": 36864
          }
        },
        {
          "id": "k80-two-boards",
          "rank": 1,
          "hardware": {
            "unit_type": "k80-gpu",
            "count": 4,
            "same_card": false,
            "per_card": 2,
            "min_mem_per_unit": 10340368384,
            "min_host_ram": 7000000000
          },
          "load": {
            "ctx_cap": 40960,
            "max_sessions": 1,
            "max_concurrent": 0,
            "reasoning_budget": 2000,
            "mtp": true
          },
          "options": {
            "kv_type": "q4",
            "ctx_min": 8192,
            "ctx_max": 131072,
            "sessions_max": 4,
            "mem": {
              "fixed": 11367743488,
              "ctx_per_token": 2230,
              "per_session": 13631488,
              "kv_per_token": 1766
            }
          },
          "sampling": {
            "temperature": 0.6,
            "top_p": 0.95,
            "top_k": 20,
            "min_p": 0,
            "presence_penalty": 0
          },
          "measured": {
            "decode_tok_s": {
              "8k": 21.3,
              "14k": 22.0,
              "32k": 13.8
            },
            "ttft": {
              "prompt_tokens": 8444,
              "seconds": 51.4
            },
            "context_window": 40960
          }
        }
      ]
    },
    {
      "id": "ornith-1.5-9b-q8",
      "name": "Ornith 1.5 9B",
      "family": "ornith-1.5",
      "description": "ornith-ai's 9B dense coding and agent model on one Tenstorrent Blackhole p150 or one Tesla K80 GPU, with MTP speculative decoding.",
      "licence": {
        "id": "mit",
        "url": "https://huggingface.co/ornith-ai/Ornith-1.5-9B"
      },
      "capabilities": [
        "chat",
        "tools",
        "reasoning"
      ],
      "files": [
        {
          "role": "weights",
          "hf_repo": "TeamIDE/Ornith-1.5-9B-Pantheon-TT",
          "revision": "86feee7f479dce675c821fad48dc77e76df1a3b5",
          "filename": "Ornith-1.5-9B-Q8_0.gguf",
          "size": 9786060384,
          "sha256": "22086870b009dbe9815ee752c48a82de930118a7c5ce5599590892ae03b8b010"
        },
        {
          "role": "extra",
          "hf_repo": "TeamIDE/Ornith-1.5-9B-Pantheon-TT",
          "revision": "86feee7f479dce675c821fad48dc77e76df1a3b5",
          "filename": "Ornith-1.5-9B-Q8_0.bfp8.fttw",
          "size": 9778719184,
          "sha256": "a81f0d06875c29997639abc5192de06cfae457891df56314a839b4111054cd79",
          "unit_types": ["p150"]
        },
        {
          "role": "extra",
          "hf_repo": "TeamIDE/Ornith-1.5-9B-Pantheon-TT",
          "revision": "86feee7f479dce675c821fad48dc77e76df1a3b5",
          "filename": "Ornith-1.5-9B-Q8_0.bfp8.mtp.fttw",
          "size": 258561216,
          "sha256": "db5ebc0d1864b29125623cd8a96c97268ec7aea919e2e52b53436732f7582bfb",
          "unit_types": ["p150"]
        }
      ],
      "profiles": [
        {
          "id": "p150-one",
          "rank": 1,
          "hardware": {
            "unit_type": "p150",
            "count": 1,
            "same_card": false,
            "min_mem_per_unit": 0,
            "min_host_ram": 7000000000
          },
          "load": {
            "ctx_cap": 131072,
            "max_sessions": 1,
            "max_concurrent": 0,
            "reasoning_budget": 0,
            "mtp": true
          },
          "options": {
            "kv_type": "bf16",
            "ctx_min": 8192,
            "ctx_max": 131072,
            "sessions_max": 1,
            "mem": {
              "fixed": 11939740647,
              "ctx_per_token": 0,
              "per_session": 0,
              "kv_per_token": 36864
            }
          },
          "sampling": {
            "temperature": 0.6,
            "top_p": 0.95,
            "top_k": 20,
            "min_p": 0,
            "presence_penalty": 0
          },
          "measured": {
            "decode_tok_s": {
              "8k": 40.8,
              "14k": 37.92,
              "32k": 0
            },
            "ttft": {
              "prompt_tokens": 7395,
              "seconds": 8.465
            },
            "context_window": 131072
          }
        },
        {
          "id": "k80-one",
          "rank": 2,
          "hardware": {
            "unit_type": "k80-gpu",
            "count": 1,
            "same_card": false,
            "min_mem_per_unit": 8982102016,
            "min_host_ram": 7000000000
          },
          "load": {
            "ctx_cap": 32768,
            "max_sessions": 1,
            "max_concurrent": 0,
            "reasoning_budget": 0,
            "mtp": true
          },
          "options": {
            "kv_type": "q4",
            "ctx_min": 8192,
            "ctx_max": 131072,
            "sessions_max": 4,
            "mem": {
              "fixed": 9895411712,
              "ctx_per_token": 10368,
              "per_session": 0,
              "kv_per_token": 9216
            }
          },
          "sampling": {
            "temperature": 0.6,
            "top_p": 0.95,
            "top_k": 20,
            "min_p": 0,
            "presence_penalty": 0
          },
          "measured": {
            "decode_tok_s": {
              "8k": 7.7,
              "14k": 8.29,
              "32k": 7.85
            },
            "ttft": {
              "prompt_tokens": 8445,
              "seconds": 424.6
            },
            "context_window": 32768
          }
        },
        {
          "id": "k80-one-fp32",
          "rank": 3,
          "hardware": {
            "unit_type": "k80-gpu",
            "count": 1,
            "same_card": false,
            "min_mem_per_unit": 9344909312,
            "min_host_ram": 7000000000
          },
          "load": {
            "ctx_cap": 16384,
            "max_sessions": 1,
            "max_concurrent": 0,
            "reasoning_budget": 0,
            "mtp": false
          },
          "options": {
            "kv_type": "fp32",
            "ctx_min": 8192,
            "ctx_max": 16384,
            "sessions_max": 4,
            "mem": {
              "fixed": 9286189056,
              "ctx_per_token": 65536,
              "per_session": 58720256,
              "kv_per_token": 65536
            }
          },
          "sampling": {
            "temperature": 0.6,
            "top_p": 0.95,
            "top_k": 20,
            "min_p": 0,
            "presence_penalty": 0
          },
          "measured": {
            "decode_tok_s": {
              "8k": 6.2,
              "14k": 5.99,
              "32k": 0
            },
            "ttft": {
              "prompt_tokens": 8445,
              "seconds": 411.5
            },
            "context_window": 16384
          }
        }
      ]
    }
  ]
}