v0.8.1

0.8.1 (2026-05-23)

Bug Fixes

• constrain cuda release architectures (#166) (8986b4c)

v0.8.0

0.8.0 (2026-05-23)

Features

• build upstream llama.cpp releases (#164) (ad9322e)

v0.7.44

0.7.44 (2026-05-22)

Bug Fixes

• cache models by gguf filename (#162) (6cacc64)

v0.7.43

0.7.43 (2026-05-22)

Bug Fixes

• omit server-only unsupported draft split flag (#160) (cc80138)

v0.7.42

0.7.42 (2026-05-22)

Bug Fixes

• align speculative flags with atomic nextn (2a95dc4)

v0.7.41

0.7.41 (2026-05-22)

Bug Fixes

• build qwen nextn turboquant llama source (de42cfb)

v0.7.40

0.7.40 (2026-05-22)

Bug Fixes

• support runner compiler bin path (#154) (e7f0816)

v0.7.39

0.7.39 (2026-05-22)

Bug Fixes

• resolve c compiler frontend path (7da9672)

v0.7.38

0.7.38 (2026-05-22)

Bug Fixes

• package release images with crane (f9bd9fa)

v0.7.35

0.7.35 (2026-05-22)

Bug Fixes

• stabilize offline native release builds (#143) (a4f32b9)
• verify baked llama source ref (#145) (e835b53)