opencoti-llamafile-0.10.3-c3-aarch64.llamafile.MANIFEST.json

{
  "artifact": "opencoti-llamafile-0.10.3-c3-aarch64.llamafile",
  "sha256": "3032ca64b7675b3aa65203425e6b76f6f4203749e223fb12e63d7d56babc44a2",
  "sizeBytes": 1945433983,
  "builtAt": "2026-07-15T10:38:35.591Z",
  "llamafileVersion": "0.10.3",
  "opencotiTag": "c3",
  "versionString": "opencoti-0.10.3-c3",
  "gitCommit": "c010fac45839449b8721ce4293c5678bf1edd254",
  "arch": "aarch64",
  "opencotiPatches": [
    "0001-lazy-slot-context-defer.patch",
    "0002-lazy-slot-context-grow.patch",
    "0003-arg-assertion-bypass.patch",
    "0004-cuda-std-cpp17.patch",
    "0005-lazy-slot-context-shrink.patch",
    "0006-kv-reuse-prefix.patch",
    "0007-build-header-deps.patch",
    "0010-rest-kv-eviction.patch",
    "0020-headinfer-per-head.patch",
    "0030-neo-pipeline.patch",
    "0031-per-stream-split.patch",
    "0032-m2-state-io.patch",
    "0033-cpy-tie-blck-align.patch",
    "0034-m2-pinned-host.patch",
    "0035-concat-q-block-aware.patch",
    "0036-pcie-probe-consume.patch",
    "0040-iqk-flash-attn.patch",
    "0042-fused-moe.patch",
    "0070-rolling-kv.patch",
    "0071-state-io-cosmocc.patch",
    "0072-poly-kv-pool.patch",
    "0073-turboquant-kv.patch",
    "0075-d512-turbo-vec.patch",
    "0078-dca.patch",
    "0079-gemma-assistant-mtp.patch",
    "0080-mtp-d512-fa.patch",
    "0081-tcq-q6-asym-kv.patch",
    "0082-cuda-build-parallelism.patch",
    "0083-tcq-warp-encode.patch",
    "0084-cuda-graphs.patch",
    "0085-dca-quant-prefill-inlaunch.patch",
    "0086-f16-moe-mmf-prefill.patch",
    "0087-dca-yarn-factor.patch",
    "0088-sparsev-autopolicy.patch",
    "0089-sparse-attn-vslash-foundation.patch",
    "0090-rolling-kv-compute-reserve.patch",
    "0091-multi-session-graph-nodes.patch",
    "0092-cuda-build-ccache.patch",
    "0093-mtp-dualctx-fused-nextn.patch",
    "0094-tinyblas-small-gemm.patch",
    "0095-prefill-checkpoint-align.patch",
    "0096-remove-single-ctx-gemma-mtp.patch",
    "0097-native-elf-cmake.patch",
    "0098-rolling-kv-lse-decode.patch",
    "0099-dca-gemma4-wiring.patch",
    "0100-dca-single-chunk-defer.patch",
    "0101-kv-attn-rot-inherit.patch",
    "0102-fattn-bulk-h2d-scratch-ring.patch",
    "0103-kv-cpu-fa-probe-guard.patch",
    "0104-fattn-q6q4-vec-instance.patch",
    "0105-ws1-dca-per-tensor-defer.patch",
    "0106-ws1-scalar-k-turbo-v-boot-correctness.patch",
    "0107-ws1-turbo-mma-fused-decode.patch",
    "0108-ws2-dca-fused-turbo-v-inregister.patch",
    "0109-cuda-dso-2gb-fatbin-last.patch",
    "0110-ws2-asym-band-nan-guard.patch",
    "0111-dca-streaming-tail-bug2144.patch",
    "0112-rolling-kv-phasegate-prefill-bug2145.patch",
    "0113-rolling-kv-reserve-fixpoint-bug2146.patch",
    "0114-hybrid-rolling-kv-residency-bug2142.patch",
    "0115-rolling-kv-shift-guard-bug2148.patch",
    "0116-kv-auto-tier-hal582-p1.patch",
    "0117-rolling-kv-shift-rerope-639.patch",
    "0118-kv-auto-tier-tstar-spill.patch",
    "0119-p0-dcaon-scalar-inregister.patch",
    "0120-p2-mixed-kv.patch",
    "0121-p2-auto-tail.patch",
    "0122-rys-layer-duplication.patch",
    "0123-rys-probe.patch",
    "0124-rys-mtp-ctx-guard-bug2171.patch",
    "0125-hybrid-window-orderagnostic-bug2172.patch",
    "0126-p0-dcaoff-scalar-mma-decode.patch",
    "0127-p0-dcaoff-scalar-mma-decode-d256-d512.patch",
    "0128-mtp-spec-clone-cheap-mmvf-verify.patch",
    "0129-rys-probe-nested-args.patch",
    "0130-parallel2-fit-ckpt-fixes.patch",
    "0131-rys-probe-task-battery-v2.patch",
    "0132-runtime-introspection.patch",
    "0133-opencoti-version-string.patch",
    "0134-cuda-cross-sbsa.patch"
  ],
  "backends": [
    {
      "backend": "cuda-sbsa",
      "dsoFilename": "ggml-cuda.so",
      "sha256": "6cb890fdf0a374acc7bfd86a158e3ccb2270e17259346e6190de0aead66937ee",
      "sizeBytes": 1889975248
    }
  ]
}