pin v2 payload: engine 2610182149001, cuda 2610160820001, cuda12 2610141455002, sbsa 2610141455003, vulkan 2610182214001, macos 2610182214002, media 2610071902001

.gitattributes

@@ -302,3 +302,8 @@ components/engine/2610161651001/opencoti-0.10.5-c13-2610161651001 filter=lfs dif
 components/engine/2610161651001/opencoti-0.10.5-c13-2610161651001.exe filter=lfs diff=lfs merge=lfs -text
 components/engine/2610171128001/opencoti-0.10.5-c13-2610171128001 filter=lfs diff=lfs merge=lfs -text
 components/engine/2610171128001/opencoti-0.10.5-c13-2610171128001.exe filter=lfs diff=lfs merge=lfs -text
+components/engine/2610182149001/opencoti-0.10.5-c13-2610182149001 filter=lfs diff=lfs merge=lfs -text
+components/engine/2610182149001/opencoti-0.10.5-c13-2610182149001.exe filter=lfs diff=lfs merge=lfs -text
+components/vulkan/2610182214001/ggml-vulkan-x86_64.so filter=lfs diff=lfs merge=lfs -text
+components/vulkan/2610182214001/ggml-vulkan-win-x86_64.dll filter=lfs diff=lfs merge=lfs -text
+components/macos/2610182214002/ggml-metal-aarch64.dylib filter=lfs diff=lfs merge=lfs -text

components/cuda/2610160820001/BUILD_INFO.cuda.md

@@ -2,7 +2,7 @@
 
 Component of the opencoti pin format v2 (docs/protocols/PIN_FORMAT.md).
 
-Built from the engine source of build 2610141255001; gated floor (engine-min) 2610141212001; gated with 2610160746001, 2610161700001, 2610170942001, 2610171128001.
+Built from the engine source of build 2610141255001; gated floor (engine-min) 2610141212001; gated with 2610160746001, 2610161700001, 2610170942001, 2610181815001, 2610182159001, 2610182149001.
 
 | file | platform | kind | sha256 | bytes |
 |---|---|---|---|---|
@@ -26,6 +26,8 @@ llama.cpp/ggml/src/ggml-backend-impl.h f9fdfe6bae61d1387593b40fc0f90ceee490c6a33
 llamafile/gpu_backend.h 4929c36c034b6589e88653bcb7a5c5a225743a07320f64c0857d92bdaf048208
 ```
 
+**Gated with the r6 engines 2610181815001 and 2610182159001 (2026-10-18), the same bytes as r5:** the glibc 2.28 image on bs2 PASS; KVarN on bs2 RTX PRO 6000; identity with r5 4 / 4.
+
 **Gated with the r5 engine 2610170942001 (2026-10-17), the same bytes as r4:** the glibc 2.28 image on bs2 PASS; the notice gates (v2 and v3), the bug-4108 reproducer and greedy identity with r4 on the Linux library (RTX PRO 6000).
 
 **Gated with the r4 engine 2610161700001 (2026-10-16), the same bytes as r3:** the glibc 2.28 image on bs2 PASS; the bug-4108 reproducer and greedy identity with r3 on the Linux library (RTX PRO 6000).

components/engine/2610182149001/BUILD_INFO.engine.md

@@ -0,0 +1,42 @@
+# engine 2610182149001
+
+Component of the opencoti pin format v2 (docs/protocols/PIN_FORMAT.md).
+
+| file | platform | kind | sha256 | bytes |
+|---|---|---|---|---|
+| `opencoti-0.10.5-c13-2610182149001` | any | bin | `af7cb0f60b41eb1810b11c29e58fe9a7705f1b8c933e2c306ba14fda44e6c377` | 146062909 |
+| `opencoti-0.10.5-c13-2610182149001.exe` | win-x86_64 | bin | `af7cb0f60b41eb1810b11c29e58fe9a7705f1b8c933e2c306ba14fda44e6c377` | 146062909 |
+
+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
+```
+
+**c13 r6 dev snapshot — the development build of the c13 r6 release, for integration testing.** Engine 2610182149001 (`opencoti-0.10.5-c13`, dev channel) is built from the same source as the c13 r6 release engine 2610182159001 (chain 0001–0717; the 7569 source and embedded-UI files the two trees share are identical by sha256). The libraries are the r6 release index's. This engine ran the glibc 2.28 route gate (Vulkan + CUDA) on bs2 and the web UI smoke in headless Chrome (v9-agentic: reasoning shown, resume 200, Stop 204); the scoped regression gates of the release ran on 2610182159001 (see the release notes).
+
+**c13 r6 (2026-10-18) — a re-release of c13.** Same tag, new engine 2610182159001 (r5: 2610170942001): the dev chain 0001–0717 — tool-call abandon / KV yield on idle / think→tool-call (0694–0696), KVarN ub rows, LSE f32 accumulation and the exact-max chunk merge (0697–0704), Vulkan native KVarN attention on by default (0705–0711), the CPU-spin fix (0712), the IQ2_S Vulkan matrix-vector (0713), the Metal record-reading KVarN attention (0714), the command line in the boot log with secrets masked (0716), and the current web UI (llama-ui b11721) with the /v1/stream query route it uses (0717). New features: kv_reservation_yield_v1, tool_call_abandon_v1, tool_call_in_thinking_v1. CUDA libraries = r5's bytes; Vulkan (Linux + Windows) and the Metal library rebuilt. Scoped gates on the r6 bytes, r5 as the control: the glibc 2.28 image on bs2 PASS (Vulkan and CUDA); identity with r5 (greedy, f16, CUDA, four needle prompts) 4 / 4 byte-identical; the bug-4132 spin reproducer (cores busy during GPU decode: Vulkan 13.75 → 1.02, CUDA 13.24 → 1.32); KVarN on bs2 RTX PRO 6000, eleven2go RTX 3090 + RX 9070 XT (Vulkan) and the Mac mini M6 (Metal): every needle found, native == materialize; the repetition-penalty window gate (device == host, with and without speculation, v9-agentic) PASS; the web UI smoke in headless Chrome PASS.

components/engine/2610182149001/opencoti-0.10.5-c13-2610182149001

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:af7cb0f60b41eb1810b11c29e58fe9a7705f1b8c933e2c306ba14fda44e6c377
+size 146062909

components/engine/2610182149001/opencoti-0.10.5-c13-2610182149001.exe

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:af7cb0f60b41eb1810b11c29e58fe9a7705f1b8c933e2c306ba14fda44e6c377
+size 146062909

components/macos/2610182214002/BUILD_INFO.macos.md

@@ -0,0 +1,37 @@
+# macos 2610182214002
+
+Component of the opencoti pin format v2 (docs/protocols/PIN_FORMAT.md).
+
+Built from the engine source of build 2610181815001; gated floor (engine-min) 2610141212001; gated with 2610181815001, 2610182159001, 2610182149001.
+
+| file | platform | kind | sha256 | bytes |
+|---|---|---|---|---|
+| `ape-macos-aarch64` | macos-aarch64 | ape | `07f550415a59415973ef877eec99a776dfa5263b06fdb1fd9cefb7b884f0f723` | 56896 |
+| `ggml-metal-aarch64.dylib` | macos-aarch64 | metal | `10111332befca174aca61f99bb140739127dd8f4c4483c83a704cdc9bfc8c7c8` | 1737648 |
+
+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): `42c6a2b3-47be-4adb-83e0-9c3d0900fd17 eea7c66d-e8a4-4860-b9da-71e9416a2805`.
+
+**c13 r6 (2026-10-18):** the Metal library rebuilt with the record-reading KVarN attention (0714) and KVarN + LSE on Metal (0701–0702); Developer ID signed, notarised (42c6a2b3…, Accepted); the loader is the r5 one (eea7c66d…). Gated on the Mac mini (M6) with the r6 engines 2610181815001 and 2610182159001: loaded from the executable directory, text == materialize, every needle found, 0 failed cells; 36k-token decode 2.1–2.3x the materialize path. One Metal architecture measured.

components/macos/2610182214002/ggml-metal-aarch64.dylib

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:10111332befca174aca61f99bb140739127dd8f4c4483c83a704cdc9bfc8c7c8
+size 1737648

components/vulkan/2610182214001/BUILD_INFO.vulkan.md

@@ -0,0 +1,29 @@
+# vulkan 2610182214001
+
+Component of the opencoti pin format v2 (docs/protocols/PIN_FORMAT.md).
+
+Built from the engine source of build 2610181815001; gated floor (engine-min) 2610141212001; gated with 2610181815001, 2610182159001, 2610182149001.
+
+| file | platform | kind | sha256 | bytes |
+|---|---|---|---|---|
+| `ggml-vulkan-x86_64.so` | x86_64 | vulkan | `b496869f9502d8a5f37f97af0749035051ca6be42b4b89138147cfdcd1fbbff9` | 62299880 |
+| `ggml-vulkan-win-x86_64.dll` | win-x86_64 | vulkan | `1db8c5faa83d4a56d23f2963a7ce524f7eff8ceeda9215f53fb1413af316b02b` | 63221943 |
+
+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
+```
+
+**c13 r6 (2026-10-18):** built from the r6 clean-room (chain 0001–0714; the ggml interface unchanged, bac5050d). Carries the Vulkan native KVarN attention (0705–0711, on by default, sized on three architectures) and the IQ2_S integer-dot matrix-vector (0713: IQ2_M on RTX 3090 34.3 → 46.4 tok/s, RX 9070 XT 37.8 → 45.2). The Windows DLL is from the clean-room (bug-4136: every bs2 DEV DLL since 0413 linked stale core objects; release lanes compile from scratch). Gated with the r6 engines 2610181815001 and 2610182159001: the glibc 2.28 image on bs2 PASS; KVarN on bs2 and on eleven2go (both cards) PASS.

components/vulkan/2610182214001/ggml-vulkan-win-x86_64.dll

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1db8c5faa83d4a56d23f2963a7ce524f7eff8ceeda9215f53fb1413af316b02b
+size 63221943

components/vulkan/2610182214001/ggml-vulkan-x86_64.so

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b496869f9502d8a5f37f97af0749035051ca6be42b4b89138147cfdcd1fbbff9
+size 62299880