noonghunna 08d4dda6f3 feat(35b-a3b): byteshape IQ4_XS ik-llama single-card preset — validated (#299) ...

Intake of @Rhonstin's #293 byteshape IQ4_XS 35B-A3B preset into the curated catalog + first-party validation on 1x 3090: verify-full/stress 8/8 (NIAH→240K), bench n=5 (narrative 113/code 129 wall TPS), 8-pack 110/150 (≈ author's 111/150), soak-continuous PASS. Registry ik-llama/byteshape-iq4xs-mtp, ⚠️ Production w/ caveats. Intake fixes vs #293: image cu13, port 8058. Credit @Rhonstin. Gate 41/41.

2026-06-02 20:15:48 +05:00

..

gemma-4-26b-a4b.yml

refactor(compose): insert <quant> layer + make the registry the single source of truth

2026-05-26 15:58:08 +00:00

gemma-4-31b.yml

feat(beellama): release v0.3.0 Q8_K_XL dual composes (experimental) + centralize image pin (#296)

2026-06-02 16:35:48 +05:00

qwen3.6-27b.yml

feat(beellama): release v0.3.0 Q8_K_XL dual composes (experimental) + centralize image pin (#296)

2026-06-02 16:35:48 +05:00

qwen3.6-35b-a3b.yml

feat(35b-a3b): byteshape IQ4_XS ik-llama single-card preset — validated (#299)

2026-06-02 20:15:48 +05:00