{
  "repository": "https://github.com/PrismML-Eng/llama.cpp",
  "embedded_commit": "f0a2b5dc9ea066780b9b410d2c0f95d675859bf7",
  "build_number": 10697,
  "provenance_limit": "Built from the embedded upstream commit with the local qwen35 patch in native/source/local-qwen35.patch; full native source tree is available upstream at that commit.",
  "license": "MIT",
  "tested_build": {
    "compiler": "GNU12.4.0",
    "architecture": "Linux x86_64",
    "cuda_architectures": "120a-real",
    "GGML_CUDA": true,
    "GGML_NATIVE": true,
    "GGML_CUDA_NCCL": true
  },
  "external_libraries": [
    "libcudart.so.13",
    "libcublas.so.13",
    "libcuda.so.1",
    "libnccl.so.2",
    "GNU C/C++ runtime"
  ],
  "bundled_files": {
    "libggml-base.so": {
      "bytes": 939104,
      "sha256": "bb2e08d6d1d286044e210fccad6b87ad56cab9a905efca8c33493637d02770e4"
    },
    "libggml.so": {
      "bytes": 55384,
      "sha256": "2b8b4cb00d91c894c432744a4e53206b033922949774b2c1fddc82690bfb6c80"
    },
    "libggml-cpu.so.0.21.0": {
      "bytes": 1421624,
      "sha256": "eed2acd145c161a1c902efe435d6879a1aa7481c1ac729b93d8911931d60fc00"
    },
    "libggml-cuda.so": {
      "bytes": 49193080,
      "sha256": "b3be257e214b5d8264a57a56185d9612024ddee7f9bd38662f7e981a596353e1"
    },
    "libggml-cpu.so.0": {
      "bytes": 1421624,
      "sha256": "eed2acd145c161a1c902efe435d6879a1aa7481c1ac729b93d8911931d60fc00"
    },
    "libllama-common.so.0": {
      "bytes": 5769016,
      "sha256": "29149691e8cdba482142e324cb404f8b52fc28addd0f1b3f7d8b164468ca860c"
    },
    "libmtmd.so.0.2.0": {
      "bytes": 1874536,
      "sha256": "022c08839b66df4d936bc7a34f0386706e01aeb07f165a6a572636d5659be97a"
    },
    "libllama-server-impl.so": {
      "bytes": 6863696,
      "sha256": "e22898b4b47edbf8576222ea6fc25841de8ba66d1089bc08984b51dae4918e2a"
    },
    "libllama-cli-impl.so": {
      "bytes": 235344,
      "sha256": "919317105262dcfda2b2a73235b27f435be63f85de672fd3ed0ac6f47c3ae69b"
    },
    "libmtmd.so": {
      "bytes": 1874536,
      "sha256": "022c08839b66df4d936bc7a34f0386706e01aeb07f165a6a572636d5659be97a"
    },
    "libggml-cuda.so.0": {
      "bytes": 49193080,
      "sha256": "b3be257e214b5d8264a57a56185d9612024ddee7f9bd38662f7e981a596353e1"
    },
    "libllama-common.so": {
      "bytes": 5769016,
      "sha256": "29149691e8cdba482142e324cb404f8b52fc28addd0f1b3f7d8b164468ca860c"
    },
    "libggml.so.0": {
      "bytes": 55384,
      "sha256": "2b8b4cb00d91c894c432744a4e53206b033922949774b2c1fddc82690bfb6c80"
    },
    "libllama.so.0.2.0": {
      "bytes": 4617376,
      "sha256": "a38adf43b9dc0e1d86620fa79d94a480f75d59733205cdd8d29281bacb397ba9"
    },
    "llama-server": {
      "bytes": 18000,
      "sha256": "9f89e204f245718c4d3efead026c0b9cedb062e416b3267a9ebaee2328f74328"
    },
    "libggml-base.so.0": {
      "bytes": 939104,
      "sha256": "bb2e08d6d1d286044e210fccad6b87ad56cab9a905efca8c33493637d02770e4"
    },
    "libggml.so.0.21.0": {
      "bytes": 55384,
      "sha256": "2b8b4cb00d91c894c432744a4e53206b033922949774b2c1fddc82690bfb6c80"
    },
    "libllama.so.0": {
      "bytes": 4617376,
      "sha256": "a38adf43b9dc0e1d86620fa79d94a480f75d59733205cdd8d29281bacb397ba9"
    },
    "libllama-common.so.0.2.0": {
      "bytes": 5769016,
      "sha256": "29149691e8cdba482142e324cb404f8b52fc28addd0f1b3f7d8b164468ca860c"
    },
    "libllama.so": {
      "bytes": 4617376,
      "sha256": "a38adf43b9dc0e1d86620fa79d94a480f75d59733205cdd8d29281bacb397ba9"
    },
    "libmtmd.so.0": {
      "bytes": 1874536,
      "sha256": "022c08839b66df4d936bc7a34f0386706e01aeb07f165a6a572636d5659be97a"
    },
    "libggml-cuda.so.0.21.0": {
      "bytes": 49193080,
      "sha256": "b3be257e214b5d8264a57a56185d9612024ddee7f9bd38662f7e981a596353e1"
    },
    "libggml-base.so.0.21.0": {
      "bytes": 939104,
      "sha256": "bb2e08d6d1d286044e210fccad6b87ad56cab9a905efca8c33493637d02770e4"
    },
    "libggml-cpu.so": {
      "bytes": 1421624,
      "sha256": "eed2acd145c161a1c902efe435d6879a1aa7481c1ac729b93d8911931d60fc00"
    }
  },
  "launch_config": {
    "context": 4096,
    "parallel": 1,
    "reasoning": false,
    "cache_prompt": false,
    "gpu_layers": 99,
    "flash_attention": true
  }
}