pin v2 payload: engine 2611010728001, cuda 2610311711001, cuda12 2610311711002, sbsa 2610311711003, vulkan 2610311711004, macos 2610311711005, media 2610071902001

.gitattributes

@@ -309,3 +309,13 @@ components/vulkan/2610182214001/ggml-vulkan-win-x86_64.dll filter=lfs diff=lfs m
 components/macos/2610182214002/ggml-metal-aarch64.dylib filter=lfs diff=lfs merge=lfs -text
 components/cuda12/2610270912001/ggml-cuda-cu12-x86_64.so filter=lfs diff=lfs merge=lfs -text
 components/cuda12/2610270912001/ggml-cuda-cu12-x86_64.dll filter=lfs diff=lfs merge=lfs -text
+components/engine/2611010728001/opencoti-0.10.5-c14-2611010728001 filter=lfs diff=lfs merge=lfs -text
+components/engine/2611010728001/opencoti-0.10.5-c14-2611010728001.exe filter=lfs diff=lfs merge=lfs -text
+components/cuda/2610311711001/ggml-cuda-x86_64.so filter=lfs diff=lfs merge=lfs -text
+components/cuda/2610311711001/ggml-cuda-win-x86_64.dll filter=lfs diff=lfs merge=lfs -text
+components/cuda12/2610311711002/ggml-cuda-cu12-x86_64.so filter=lfs diff=lfs merge=lfs -text
+components/cuda12/2610311711002/ggml-cuda-cu12-win-x86_64.dll filter=lfs diff=lfs merge=lfs -text
+components/sbsa/2610311711003/ggml-cuda-sbsa-aarch64.so filter=lfs diff=lfs merge=lfs -text
+components/vulkan/2610311711004/ggml-vulkan-x86_64.so filter=lfs diff=lfs merge=lfs -text
+components/vulkan/2610311711004/ggml-vulkan-win-x86_64.dll filter=lfs diff=lfs merge=lfs -text
+components/macos/2610311711005/ggml-metal-aarch64.dylib filter=lfs diff=lfs merge=lfs -text

components/cuda/2610311711001/BUILD_INFO.cuda.md

@@ -0,0 +1,29 @@
+# cuda 2610311711001
+
+Component of the opencoti pin format v2 (docs/protocols/PIN_FORMAT.md).
+
+Built from the engine source of build 2610312243001; gated floor (engine-min) 2610312243001; gated with 2610312243001, 2611010728001.
+
+| file | platform | kind | sha256 | bytes |
+|---|---|---|---|---|
+| `ggml-cuda-x86_64.so` | x86_64 | cuda | `f28c05ff0d14986c645f0a5152d8cfe6cefa2e10edcbdb200c559bbac432973e` | 792419568 |
+| `ggml-cuda-win-x86_64.dll` | win-x86_64 | cuda | `14d6564cddba72f007b036c5ed742aad417a6ea9cfbcc694f32837c3a44267bb` | 767183872 |
+
+ABI `ggml` = `bac5050ddc8c58c51f0b059a7f9753b33af0ed79d355090e7c9a17ffaaaf340a` (exported) — sha256 of this list:
+
+```
+llama.cpp/ggml/include/ggml-alloc.h 94e4cd069b9313b2ceb35dacec901981e0bb478d8bb31035b7126be091998c23
+llama.cpp/ggml/include/ggml-backend.h 764d1640ad62579513c4837539708f89e8603c9dbeae053a13d1cbd84a82c865
+llama.cpp/ggml/include/ggml-cuda.h 674ea064021cf5b9970e78f1209e513dd23350b94a44a705a27eec1423fe4b0a
+llama.cpp/ggml/include/ggml-metal.h 322f36cd30f3e9e7aad7b5b9bc63078012fd0b7706ac4e24f5721b604f3d8980
+llama.cpp/ggml/include/ggml-neo-pipeline.h 190f39e183542d16588a3104d8cb4359bf52fc7a62afe8d8cc4c822f4085086d
+llama.cpp/ggml/include/ggml-oc-kvarn-range.h 76635fdd8978632f670640ddf5125a666441e3d47c680bb8b5c0a099e4861202
+llama.cpp/ggml/include/ggml-oc-ledger.h c1be11d69a57a1e721f1ae8720b87887d239ac0183d4277fbaa07f1be257d20c
+llama.cpp/ggml/include/ggml-vulkan.h f9df4f7161a84ce7600d53c3fac8a62837d5716676c9f38cd340699ef1c070d9
+llama.cpp/ggml/include/ggml.h ecfbf114f9e9151d22e30ef6d5139570c5c41b595843ca4f99125bd0196709aa
+llama.cpp/ggml/include/opencoti-devtype.h 94dea946173449bb76bffd7f733ab822fe658fe7198992de750ac012d29308ec
+llama.cpp/ggml/src/ggml-backend-impl.h f9fdfe6bae61d1387593b40fc0f90ceee490c6a339ec6d7388c51e4fca318d8f
+llamafile/gpu_backend.h 4929c36c034b6589e88653bcb7a5c5a225743a07320f64c0857d92bdaf048208
+```
+
+**c14 (2026-10-31):** CUDA 13, six generations (sm_75, 80, 86, 89, 90, 120; `120` is the sm_120f kernel set), built from the c14 clean-room on a Linode runner (the whole set compiled: `common.cuh` and `fattn-common.cuh` changed, 405 of 405 units, `OPENCOTI_ALLOW_MASS_REBUILD=1` told to the owner); the Windows DLL from the same sources on the runner's wine lane shipped from bs2 (`SHIP-OK`, 408 of 408 units, `GGML_CUDA_FA_ALL_QUANTS`). Carries: every PDL-launched kernel waits for its producer (0750, bug-4183 — a kernel reading another's output without the sync decoded garbage on Blackwell), the KVarN decode split over live cells and its item-sized partials (0725, 0729, 0732), the record payload staged in shared memory (0737), the PolyKV shared pass (0764–0769), the native KVarN MMA record stage and word unpack (0766, 0767), the row-chunk KQ mask (0785, 0789), PrismML's PQ2_0 / PTQ1_0 mat-vec kernels (0747–0749), `-sm tensor` for qwen35 (0752), the SSM_CONV + SILU fusion range check (0787), the CUDA graph changes (0776, 0778, 0779, 0782), the KVarN chunk split (0775). <gates at the publish>

components/cuda/2610311711001/ggml-cuda-win-x86_64.dll

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:14d6564cddba72f007b036c5ed742aad417a6ea9cfbcc694f32837c3a44267bb
+size 767183872

components/cuda/2610311711001/ggml-cuda-x86_64.so

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f28c05ff0d14986c645f0a5152d8cfe6cefa2e10edcbdb200c559bbac432973e
+size 792419568

components/cuda12/2610311711002/BUILD_INFO.cuda12.md

@@ -0,0 +1,29 @@
+# cuda12 2610311711002
+
+Component of the opencoti pin format v2 (docs/protocols/PIN_FORMAT.md).
+
+Built from the engine source of build 2610312243001; gated floor (engine-min) 2610312243001; gated with 2611010728001.
+
+| file | platform | kind | sha256 | bytes |
+|---|---|---|---|---|
+| `ggml-cuda-cu12-x86_64.so` | x86_64 | cuda12 | `5c4e5a8b81e8e24c7378ce5f943cb854202334f44bbe61b478077becf0bb267e` | 273138632 |
+| `ggml-cuda-cu12-win-x86_64.dll` | win-x86_64 | cuda12 | `5a698b44b536ad87b125d4076c814d674250640100328eea724c68c02a9529ac` | 256217600 |
+
+ABI `ggml` = `bac5050ddc8c58c51f0b059a7f9753b33af0ed79d355090e7c9a17ffaaaf340a` (exported) — sha256 of this list:
+
+```
+llama.cpp/ggml/include/ggml-alloc.h 94e4cd069b9313b2ceb35dacec901981e0bb478d8bb31035b7126be091998c23
+llama.cpp/ggml/include/ggml-backend.h 764d1640ad62579513c4837539708f89e8603c9dbeae053a13d1cbd84a82c865
+llama.cpp/ggml/include/ggml-cuda.h 674ea064021cf5b9970e78f1209e513dd23350b94a44a705a27eec1423fe4b0a
+llama.cpp/ggml/include/ggml-metal.h 322f36cd30f3e9e7aad7b5b9bc63078012fd0b7706ac4e24f5721b604f3d8980
+llama.cpp/ggml/include/ggml-neo-pipeline.h 190f39e183542d16588a3104d8cb4359bf52fc7a62afe8d8cc4c822f4085086d
+llama.cpp/ggml/include/ggml-oc-kvarn-range.h 76635fdd8978632f670640ddf5125a666441e3d47c680bb8b5c0a099e4861202
+llama.cpp/ggml/include/ggml-oc-ledger.h c1be11d69a57a1e721f1ae8720b87887d239ac0183d4277fbaa07f1be257d20c
+llama.cpp/ggml/include/ggml-vulkan.h f9df4f7161a84ce7600d53c3fac8a62837d5716676c9f38cd340699ef1c070d9
+llama.cpp/ggml/include/ggml.h ecfbf114f9e9151d22e30ef6d5139570c5c41b595843ca4f99125bd0196709aa
+llama.cpp/ggml/include/opencoti-devtype.h 94dea946173449bb76bffd7f733ab822fe658fe7198992de750ac012d29308ec
+llama.cpp/ggml/src/ggml-backend-impl.h f9fdfe6bae61d1387593b40fc0f90ceee490c6a339ec6d7388c51e4fca318d8f
+llamafile/gpu_backend.h 4929c36c034b6589e88653bcb7a5c5a225743a07320f64c0857d92bdaf048208
+```
+
+**c14 (2026-10-31):** CUDA 12.8, for GPUs older than sm_75 (sm_52, 61, 70), Linux library from the c14 clean-room on a Linode runner (405 of 405 units) and the Windows DLL from bs2's wine lane (CUDA 12.8.1 Windows toolkit, TinyBLAS + the CUTLASS sm_70 GEMM, 408 of 408 units). The engine takes it on Linux and Windows when every NVIDIA GPU is below 7.5, and from c14 REFUSES the CUDA 13 library on such a GPU when this one is missing (0792, bug-4247). <gates at the publish: GTX 1070 (zns, Windows) or stated as not run>

components/cuda12/2610311711002/ggml-cuda-cu12-win-x86_64.dll

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5a698b44b536ad87b125d4076c814d674250640100328eea724c68c02a9529ac
+size 256217600

components/cuda12/2610311711002/ggml-cuda-cu12-x86_64.so

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5c4e5a8b81e8e24c7378ce5f943cb854202334f44bbe61b478077becf0bb267e
+size 273138632

components/engine/2611010728001/BUILD_INFO.engine.md

@@ -0,0 +1,43 @@
+# engine 2611010728001
+
+Component of the opencoti pin format v2 (docs/protocols/PIN_FORMAT.md).
+
+| file | platform | kind | sha256 | bytes |
+|---|---|---|---|---|
+| `opencoti-0.10.5-c14-2611010728001` | any | bin | `cb40345b69eb00e07d3c26f7d36fcbcd2a911d38f67f0f00c5a5ca77af31f80b` | 146797975 |
+| `opencoti-0.10.5-c14-2611010728001.exe` | win-x86_64 | bin | `cb40345b69eb00e07d3c26f7d36fcbcd2a911d38f67f0f00c5a5ca77af31f80b` | 146797975 |
+
+ABI `ggml` = `bac5050ddc8c58c51f0b059a7f9753b33af0ed79d355090e7c9a17ffaaaf340a` (exported) — sha256 of this list:
+
+```
+llama.cpp/ggml/include/ggml-alloc.h 94e4cd069b9313b2ceb35dacec901981e0bb478d8bb31035b7126be091998c23
+llama.cpp/ggml/include/ggml-backend.h 764d1640ad62579513c4837539708f89e8603c9dbeae053a13d1cbd84a82c865
+llama.cpp/ggml/include/ggml-cuda.h 674ea064021cf5b9970e78f1209e513dd23350b94a44a705a27eec1423fe4b0a
+llama.cpp/ggml/include/ggml-metal.h 322f36cd30f3e9e7aad7b5b9bc63078012fd0b7706ac4e24f5721b604f3d8980
+llama.cpp/ggml/include/ggml-neo-pipeline.h 190f39e183542d16588a3104d8cb4359bf52fc7a62afe8d8cc4c822f4085086d
+llama.cpp/ggml/include/ggml-oc-kvarn-range.h 76635fdd8978632f670640ddf5125a666441e3d47c680bb8b5c0a099e4861202
+llama.cpp/ggml/include/ggml-oc-ledger.h c1be11d69a57a1e721f1ae8720b87887d239ac0183d4277fbaa07f1be257d20c
+llama.cpp/ggml/include/ggml-vulkan.h f9df4f7161a84ce7600d53c3fac8a62837d5716676c9f38cd340699ef1c070d9
+llama.cpp/ggml/include/ggml.h ecfbf114f9e9151d22e30ef6d5139570c5c41b595843ca4f99125bd0196709aa
+llama.cpp/ggml/include/opencoti-devtype.h 94dea946173449bb76bffd7f733ab822fe658fe7198992de750ac012d29308ec
+llama.cpp/ggml/src/ggml-backend-impl.h f9fdfe6bae61d1387593b40fc0f90ceee490c6a339ec6d7388c51e4fca318d8f
+llamafile/gpu_backend.h 4929c36c034b6589e88653bcb7a5c5a225743a07320f64c0857d92bdaf048208
+```
+
+ABI `media` = `b099409b3029a6e380268e31d1caefa6706569961d32e72b3543fbc73903d95c` (exported) — sha256 of this list:
+
+```
+llamafile/oc-audiocpp/audiocpp.h 6ad0b55d5f28adc9ad60a45bf9a7caa428c4f5441389a6932d6670a8e2a90d73
+llamafile/oc-codec/oc_codec_sidecar.h f2c9001d1414aa2a55533b867b79c52a52c682558e0234c5bebd30bf97b91005
+```
+
+ABI `loader` = `d6ff34022f1d3fb5af56ca0c80e3f4e552bf09f1bebdf0cb5f314a454bf38dbc` (exported) — sha256 of this list:
+
+```
+.cosmocc/4.0.2/bin/ape-m1.c 78af79e20abbb3f99a97355959b59bd58949564064adb754f0302ebabb4bf7a3
+```
+
+**c14 dev snapshot — the development build of the c14 release, for integration testing.** Engine 2611010728001 (`opencoti-0.10.5-c14`, dev channel) is built from the same source as the c14 release engine 2610312243001 (the 12335 source files the two trees share are identical by sha256 (0 differ; clean-room /mnt/sdc/opencoti-cleanroom-c14 at 132534bb26)). The libraries are the c14 release index's. This engine ran the glibc 2.28 route gate (Vulkan + CUDA) on bs2 with the release libraries (cuda, cuda12, vulkan); the release gates ran on 2610312243001 (see the release notes).
+
+**c14 (2026-10-31) — a new cut.** `--version`: `llamafile v0.10.5 (opencoti-0.10.5-c14) build 2610312243001`; chain 0001–0822 (c13 r6 ended at 0717). Built in the clean-room from `release/c14`. What changes: `GET /polykv/pools` (and `/kv`, `/elastic`) shows a pool op at once — create, fork, release, pin, admission force the read snapshot (0822, bug-4304; the list could stay stale on an idle engine since c13); no admission 429 without PolyKV (0718, bug-4140); the PolyKV gate books what it admits and refuses from a measured law (0721, 0724); the admission law `t = W(total rows) + k D` with a composition-matched k and one accounting for the PolyKV and elastic gates (0773, 0777, 0812, 0818); prefill and decode steps alternate by default so one prompt never stalls the decoders for its whole length (`--pd-step-budget alternate`, 0722, 0815) with the opt-in bounded prompt step `--pd-prefill-step-ratio` (0819); a session's step reads only its own live cells on CUDA, Vulkan and Metal (0725–0731, 0762), 256-cell KV pages by default (0770); PolyKV decode reads each shared block once for all its sessions (0764–0769, −23 % at 16 sessions); the KQ mask built on the device in row chunks (0785–0789); run-time KV commits that never fail (0813) and the Windows commit guard (0814); the lazy / split pinned KVarN host arena (0756–0758); `--moe-strategy` (0739, 0740); the drafter changes (0754, 0772, 0791–0794, 0806–0808: `--auto-mtp-policy adaptive`, opt-in); the CUDA 12 legacy refusal (0792, bug-4247); LFM3 and LFM3-VL (0817); the outer llamafile 0.10.6 commits (0816); the xollama #26 fixes (0738). Features: the c13 r6 set, no new names. <gates at the publish>
+Feature rows: c13 r6's 73 plus four pin-only rows for xollama's capability gate (mails #2041/#2042, 2026-10-31): `arch_lfm3_v1` (0817), `vulkan_matvec_tune_v1` (0802/0803/0810, opt-in), `auto_mtp_adaptive_v1` (0807/0808, opt-in), `moe_strategy_v1` (0739/0740). /props `features` of this engine does not list them (unchanged from c13 r6); the pin is the capability source for consumers.

components/engine/2611010728001/opencoti-0.10.5-c14-2611010728001

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cb40345b69eb00e07d3c26f7d36fcbcd2a911d38f67f0f00c5a5ca77af31f80b
+size 146797975

components/engine/2611010728001/opencoti-0.10.5-c14-2611010728001.exe

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cb40345b69eb00e07d3c26f7d36fcbcd2a911d38f67f0f00c5a5ca77af31f80b
+size 146797975

components/macos/2610311711005/BUILD_INFO.macos.md

@@ -0,0 +1,37 @@
+# macos 2610311711005
+
+Component of the opencoti pin format v2 (docs/protocols/PIN_FORMAT.md).
+
+Built from the engine source of build 2610312243001; gated floor (engine-min) 2610312243001; gated with 2610312243001.
+
+| file | platform | kind | sha256 | bytes |
+|---|---|---|---|---|
+| `ape-macos-aarch64` | macos-aarch64 | ape | `07f550415a59415973ef877eec99a776dfa5263b06fdb1fd9cefb7b884f0f723` | 56896 |
+| `ggml-metal-aarch64.dylib` | macos-aarch64 | metal | `9ae723efa04dc383138cc6a4ec8e7e8203081a9c5e64c67b317b18e2d9baa036` | 1821904 |
+
+ABI `ggml` = `bac5050ddc8c58c51f0b059a7f9753b33af0ed79d355090e7c9a17ffaaaf340a` (exported) — sha256 of this list:
+
+```
+llama.cpp/ggml/include/ggml-alloc.h 94e4cd069b9313b2ceb35dacec901981e0bb478d8bb31035b7126be091998c23
+llama.cpp/ggml/include/ggml-backend.h 764d1640ad62579513c4837539708f89e8603c9dbeae053a13d1cbd84a82c865
+llama.cpp/ggml/include/ggml-cuda.h 674ea064021cf5b9970e78f1209e513dd23350b94a44a705a27eec1423fe4b0a
+llama.cpp/ggml/include/ggml-metal.h 322f36cd30f3e9e7aad7b5b9bc63078012fd0b7706ac4e24f5721b604f3d8980
+llama.cpp/ggml/include/ggml-neo-pipeline.h 190f39e183542d16588a3104d8cb4359bf52fc7a62afe8d8cc4c822f4085086d
+llama.cpp/ggml/include/ggml-oc-kvarn-range.h 76635fdd8978632f670640ddf5125a666441e3d47c680bb8b5c0a099e4861202
+llama.cpp/ggml/include/ggml-oc-ledger.h c1be11d69a57a1e721f1ae8720b87887d239ac0183d4277fbaa07f1be257d20c
+llama.cpp/ggml/include/ggml-vulkan.h f9df4f7161a84ce7600d53c3fac8a62837d5716676c9f38cd340699ef1c070d9
+llama.cpp/ggml/include/ggml.h ecfbf114f9e9151d22e30ef6d5139570c5c41b595843ca4f99125bd0196709aa
+llama.cpp/ggml/include/opencoti-devtype.h 94dea946173449bb76bffd7f733ab822fe658fe7198992de750ac012d29308ec
+llama.cpp/ggml/src/ggml-backend-impl.h f9fdfe6bae61d1387593b40fc0f90ceee490c6a339ec6d7388c51e4fca318d8f
+llamafile/gpu_backend.h 4929c36c034b6589e88653bcb7a5c5a225743a07320f64c0857d92bdaf048208
+```
+
+ABI `loader` = `d6ff34022f1d3fb5af56ca0c80e3f4e552bf09f1bebdf0cb5f314a454bf38dbc` (exported) — sha256 of this list:
+
+```
+.cosmocc/4.0.2/bin/ape-m1.c 78af79e20abbb3f99a97355959b59bd58949564064adb754f0302ebabb4bf7a3
+```
+
+Apple notarisation (Accepted), submission id(s): `36ab297c-b099-4de4-9d2d-f96efaf9ef22 eea7c66d-e8a4-4860-b9da-71e9416a2805`.
+
+**c14 (2026-10-31):** the Metal library rebuilt from the c14 clean-room (chain 0001–0821): the native KVarN attention walks each row's live tiles under `-kvu` (0762), the KVarN decoder inside the Metal flash-attention vec kernel (0763), the device KQ mask builder (0788), the library logs at the `-lv` minimum (0761). Developer ID signed, notarised (<submission id>); the loader is the c13 r5 one (eea7c66d…, unchanged). Gated on the Mac mini (M6, macOS 27.0.1) with the c14 engine 2610312243001 (the re-cut on 0822), loaded from the executable directory: f16 generation; Qwen3.5-2B kvarn3 native == materialize at a ~6k-token needle (texts identical, needle found in both). One Metal architecture measured.

components/macos/2610311711005/ggml-metal-aarch64.dylib

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9ae723efa04dc383138cc6a4ec8e7e8203081a9c5e64c67b317b18e2d9baa036
+size 1821904

components/sbsa/2610311711003/BUILD_INFO.sbsa.md

@@ -0,0 +1,28 @@
+# sbsa 2610311711003
+
+Component of the opencoti pin format v2 (docs/protocols/PIN_FORMAT.md).
+
+Built from the engine source of build 2610312243001; gated floor (engine-min) 2610312243001; gated with nothing — NOT run.
+
+| file | platform | kind | sha256 | bytes |
+|---|---|---|---|---|
+| `ggml-cuda-sbsa-aarch64.so` | aarch64 | cuda | `14ffac18842d5a6d340380fb2b72f9a54ce45c3d7d9bbac26d5c2f27c3fbd775` | 355192632 |
+
+ABI `ggml` = `bac5050ddc8c58c51f0b059a7f9753b33af0ed79d355090e7c9a17ffaaaf340a` (exported) — sha256 of this list:
+
+```
+llama.cpp/ggml/include/ggml-alloc.h 94e4cd069b9313b2ceb35dacec901981e0bb478d8bb31035b7126be091998c23
+llama.cpp/ggml/include/ggml-backend.h 764d1640ad62579513c4837539708f89e8603c9dbeae053a13d1cbd84a82c865
+llama.cpp/ggml/include/ggml-cuda.h 674ea064021cf5b9970e78f1209e513dd23350b94a44a705a27eec1423fe4b0a
+llama.cpp/ggml/include/ggml-metal.h 322f36cd30f3e9e7aad7b5b9bc63078012fd0b7706ac4e24f5721b604f3d8980
+llama.cpp/ggml/include/ggml-neo-pipeline.h 190f39e183542d16588a3104d8cb4359bf52fc7a62afe8d8cc4c822f4085086d
+llama.cpp/ggml/include/ggml-oc-kvarn-range.h 76635fdd8978632f670640ddf5125a666441e3d47c680bb8b5c0a099e4861202
+llama.cpp/ggml/include/ggml-oc-ledger.h c1be11d69a57a1e721f1ae8720b87887d239ac0183d4277fbaa07f1be257d20c
+llama.cpp/ggml/include/ggml-vulkan.h f9df4f7161a84ce7600d53c3fac8a62837d5716676c9f38cd340699ef1c070d9
+llama.cpp/ggml/include/ggml.h ecfbf114f9e9151d22e30ef6d5139570c5c41b595843ca4f99125bd0196709aa
+llama.cpp/ggml/include/opencoti-devtype.h 94dea946173449bb76bffd7f733ab822fe658fe7198992de750ac012d29308ec
+llama.cpp/ggml/src/ggml-backend-impl.h f9fdfe6bae61d1387593b40fc0f90ceee490c6a339ec6d7388c51e4fca318d8f
+llamafile/gpu_backend.h 4929c36c034b6589e88653bcb7a5c5a225743a07320f64c0857d92bdaf048208
+```
+
+**c14 (2026-10-31):** CUDA 13 for Linux aarch64 (SBSA: DGX Spark / GB10 = sm_121, Jetson Thor = sm_110), from the c14 clean-room on a Linode runner (the whole set compiled). **Never run: there is no such machine in the test fleet.**

components/sbsa/2610311711003/ggml-cuda-sbsa-aarch64.so

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:14ffac18842d5a6d340380fb2b72f9a54ce45c3d7d9bbac26d5c2f27c3fbd775
+size 355192632

components/vulkan/2610311711004/BUILD_INFO.vulkan.md

@@ -0,0 +1,29 @@
+# vulkan 2610311711004
+
+Component of the opencoti pin format v2 (docs/protocols/PIN_FORMAT.md).
+
+Built from the engine source of build 2610312243001; gated floor (engine-min) 2610312243001; gated with 2610312243001, 2611010728001.
+
+| file | platform | kind | sha256 | bytes |
+|---|---|---|---|---|
+| `ggml-vulkan-x86_64.so` | x86_64 | vulkan | `aabaf8a7e26d51dc86a1dfac5fc2bcd5d983034b2549655ef6eb3ac50a93092b` | 68682856 |
+| `ggml-vulkan-win-x86_64.dll` | win-x86_64 | vulkan | `957802ab0856898c18e7a3bff899637e11b6f3a016677620319570b52b8a015f` | 69736207 |
+
+ABI `ggml` = `bac5050ddc8c58c51f0b059a7f9753b33af0ed79d355090e7c9a17ffaaaf340a` (exported) — sha256 of this list:
+
+```
+llama.cpp/ggml/include/ggml-alloc.h 94e4cd069b9313b2ceb35dacec901981e0bb478d8bb31035b7126be091998c23
+llama.cpp/ggml/include/ggml-backend.h 764d1640ad62579513c4837539708f89e8603c9dbeae053a13d1cbd84a82c865
+llama.cpp/ggml/include/ggml-cuda.h 674ea064021cf5b9970e78f1209e513dd23350b94a44a705a27eec1423fe4b0a
+llama.cpp/ggml/include/ggml-metal.h 322f36cd30f3e9e7aad7b5b9bc63078012fd0b7706ac4e24f5721b604f3d8980
+llama.cpp/ggml/include/ggml-neo-pipeline.h 190f39e183542d16588a3104d8cb4359bf52fc7a62afe8d8cc4c822f4085086d
+llama.cpp/ggml/include/ggml-oc-kvarn-range.h 76635fdd8978632f670640ddf5125a666441e3d47c680bb8b5c0a099e4861202
+llama.cpp/ggml/include/ggml-oc-ledger.h c1be11d69a57a1e721f1ae8720b87887d239ac0183d4277fbaa07f1be257d20c
+llama.cpp/ggml/include/ggml-vulkan.h f9df4f7161a84ce7600d53c3fac8a62837d5716676c9f38cd340699ef1c070d9
+llama.cpp/ggml/include/ggml.h ecfbf114f9e9151d22e30ef6d5139570c5c41b595843ca4f99125bd0196709aa
+llama.cpp/ggml/include/opencoti-devtype.h 94dea946173449bb76bffd7f733ab822fe658fe7198992de750ac012d29308ec
+llama.cpp/ggml/src/ggml-backend-impl.h f9fdfe6bae61d1387593b40fc0f90ceee490c6a339ec6d7388c51e4fca318d8f
+llamafile/gpu_backend.h 4929c36c034b6589e88653bcb7a5c5a225743a07320f64c0857d92bdaf048208
+```
+
+**c14 (2026-10-31):** built from the c14 clean-room on bs2 (the Linux library in the glibc 2.28 lane, floor GLIBC_2.27; the Windows DLL cross-built in the mingw lane). Carries: the KVarN store in its parallel (workspace) form (0741) with the RX 9070 XT's flush tile in device memory (0800); the V pass staging a tile's record words (0742); the scalar flash attention at head dim >= 256 on RDNA (0743: RX 9070 XT gemma-4 A4B f16 16k prefill 746 -> 2424 tok/s); the head-512 coopmat1 FA in two 32-lane subgroups and the Q split (0781, 0799); the native KVarN attention and the plain FA over each row's live cells (0726, 0728, 0730, 0731); the multi-column mat-vec for every weight type (0797), PrismML's PTQ1_0 mat-vec and signed FWHT fusion (0745, 0746), Bonsai 2 coopmat2 / integer-dot mat-mat (0744), IQ4_NL from 3 columns on NVIDIA (0805), split-k for narrow outputs (0784); the opt-in mat-vec pick `--vulkan-matvec-tune` (0802, 0803, 0810); no pipeline compile on the decode path (0809: bs2 cold 16k decode 5.1 -> 99.0 tok/s) and the Windows pipeline cache kept (0811); the row-chunk KQ mask on the device (0786); KV layers larger than the device's maximum buffer held in segments (0759); past-window KVarN record prefetch (0755); the teardown staging cycle (0804); the record-reading coopmat1 KVarN FA (0798, opt-in). Gated with the c14 engine 2610312243001 (the re-cut on 0822): the RADV iGPU rig (solidPC 5600G, RADV RENOIR, image opencoti-vulkan-radv) on the STAGED bytes beside the engine, the c13 r6 library as the other arm: f16 and kvarn3 cells, every needle found, 36/36 layers on Vulkan0, tg 30.8 / 28.1 tok/s (c13 r6: 32.4 / 28.6); the flash-attention variant set from `nm` scalar 10 / cm1 9 / cm2 3 = 22, a superset of c13 r6's 10 / 3 / 3 (the coopmat1 KVarN FA). <old-glibc, Bonsai 2 gate parts on bs2, eleven2go RTX 3090 + RX 9070 XT at the publish>

components/vulkan/2610311711004/ggml-vulkan-win-x86_64.dll

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:957802ab0856898c18e7a3bff899637e11b6f3a016677620319570b52b8a015f
+size 69736207

components/vulkan/2610311711004/ggml-vulkan-x86_64.so

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aabaf8a7e26d51dc86a1dfac5fc2bcd5d983034b2549655ef6eb3ac50a93092b
+size 68682856