GeezerVison

Commit history

1. Commits on Apr 21, 2026

   • pillar1.5(R10) ★: v0.26.0 — L2-norm formulation matches ggml exactly

     Show description for 7834d29
     

     unamedkr

     and

     claude

     committed
     7834d29View commit details

     Copy full SHA for 7834d29

     View code at this pointBrowse repository at this point
2. Commits on Apr 20, 2026

   • pillar1.5(R9): v0.25.0 — Qwen3.6 auto-serial quality mode

     Show description for a07631f
     

     unamedkr

     and

     claude

     committed
     a07631fView commit details

     Copy full SHA for a07631f

     View code at this pointBrowse repository at this point

   • pillar1.5(R8): v0.24.0 — MoE SwiGLU exact expf (Qwen3.6 coherence margin)

     Show description for 1b3575c
     

     unamedkr

     and

     claude

     committed
     1b3575cView commit details

     Copy full SHA for 1b3575c

     View code at this pointBrowse repository at this point

   • pillar1.5(R7) ★★: v0.23.0 — prompt buffer silent truncation FIXED

     Show description for a7f7b18
     

     unamedkr

     and

     claude

     committed
     a7f7b18View commit details

     Copy full SHA for a7f7b18

     View code at this pointBrowse repository at this point

   • pillar1.5(R6): v0.22.0 — chunked batched prefill for Qwen3.6 MoE hybrid

     Show description for ce3c146
     

     unamedkr

     and

     claude

     committed
     ce3c146View commit details

     Copy full SHA for ce3c146

     View code at this pointBrowse repository at this point

   • pillar1.5(R5) ★★★: v0.21.0 — Qwen3.6-35B practically usable

     Show description for 0e92d42
     

     unamedkr

     and

     claude

     committed
     0e92d42View commit details

     Copy full SHA for 0e92d42

     View code at this pointBrowse repository at this point

   • pillar1.5(R4): v0.20.0 release — NEOX RoPE + Qwen3 QK-norm root causes

     Show description for 2d0401a
     

     unamedkr

     and

     claude

     committed
     2d0401aView commit details

     Copy full SHA for 2d0401a

     View code at this pointBrowse repository at this point

   • pillar1(R6): v0.19.0 release — BPE root-cause fix

     Show description for 1e0c3d7
     

     unamedkr

     and

     claude

     committed
     1e0c3d7View commit details

     Copy full SHA for 1e0c3d7

     View code at this pointBrowse repository at this point
3. Commits on Apr 19, 2026

   • grow(R54): v0.18.0 release — Daily-Driver UX consolidation

     Show description for b061e28
     

     unamedkr

     and

     claude

     committed
     b061e28View commit details

     Copy full SHA for b061e28

     View code at this pointBrowse repository at this point

   • grow: Round 41-44 — v0.17.0 안정화 + 배포 준비 완료

     Show description for 399d061
     

     unamedkr

     and

     claude

     committed
     399d061View commit details

     Copy full SHA for 399d061

     View code at this pointBrowse repository at this point
4. Commits on Apr 13, 2026

   • feat(default): switch to Qwen3-4B — 2.4x faster + best quality

     Show description for 6ea3215
     

     unamedkr

     and

     claude

     committed
     6ea3215View commit details

     Copy full SHA for 6ea3215

     View code at this pointBrowse repository at this point

   • feat: quantcpp recommend + ask_verified Python API + CI smoke test

     Show description for 393bdd5
     

     unamedkr

     and

     claude

     committed
     393bdd5View commit details

     Copy full SHA for 393bdd5

     View code at this pointBrowse repository at this point
5. Commits on Apr 12, 2026

   • chore: bump version to v0.13.0 + release notes (#81)

     Show description for b60ce4e
     

     unamedkr

     and

     claude

     authored
     b60ce4eView commit details

     Copy full SHA for b60ce4e

     View code at this pointBrowse repository at this point

   • fix: Phi-3 Q8_0 default + unified server in CLI + CMake (#80)

     Show description for 72e815b
     

     unamedkr

     and

     claude

     authored
     72e815bView commit details

     Copy full SHA for 72e815b

     View code at this pointBrowse repository at this point

   • fix(python): add model aliases to API + graceful port conflict handling

     Show description for 6bf895b
     

     unamedkr

     and

     claude

     committed
     6bf895bView commit details

     Copy full SHA for 6bf895b

     View code at this pointBrowse repository at this point

   • feat(libturboquant): port Phi-3 fused QKV/FFN + LongRoPE to split sources

     Show description for 08e8661
     

     unamedkr

     and

     claude

     committed
     08e8661View commit details

     Copy full SHA for 08e8661

     View code at this pointBrowse repository at this point

   • feat(default): make Phi-3.5-mini the recommended default model (#66)

     Show description for eb4f7d1
     

     unamedkr

     and

     claude

     authored
     eb4f7d1View commit details

     Copy full SHA for eb4f7d1

     View code at this pointBrowse repository at this point

   • feat(phi3): end-to-end Phi-3 / Phi-3.5 architecture support (#65)

     Show description for 1e1ea2c
     

     unamedkr

     and

     claude

     authored
     1e1ea2cView commit details

     Copy full SHA for 1e1ea2c

     View code at this pointBrowse repository at this point

   • feat(feedback-quick-wins): act on 2026-04-12 external user report (#59)

     Show description for a7795a5
     

     unamedkr

     and

     claude

     authored
     a7795a5View commit details

     Copy full SHA for a7795a5

     View code at this pointBrowse repository at this point

   • fix(chat-cache): second audit pass — 9 more bugs eliminated (#53)

     Show description for 0623b16
     

     unamedkr

     and

     claude

     authored
     0623b16View commit details

     Copy full SHA for 0623b16

     View code at this pointBrowse repository at this point
6. Commits on Apr 11, 2026

   • feat: chat-mode KV cache reuse — O(history^2) → O(new tokens) per turn (#48)

     Show description for ee048f7
     

     unamedkr

     and

     claude

     authored
     ee048f7View commit details

     Copy full SHA for ee048f7

     View code at this pointBrowse repository at this point

   • feat(cli): quantcpp client (SSE streaming) + serve discoverability (#47)

     Show description for 9b8fc6e
     

     unamedkr

     and

     claude

     authored
     9b8fc6eView commit details

     Copy full SHA for 9b8fc6e

     View code at this pointBrowse repository at this point

   • chore: bump __version__ fallback to 0.12.0 (#44)

     Show description for 97a422c
     

     unamedkr

     and

     claude

     authored
     97a422cView commit details

     Copy full SHA for 97a422c

     View code at this pointBrowse repository at this point

   • feat(pypi): quantcpp CLI ollama-parity (pull/list/run/serve) (#43)

     Show description for 79db43a
     

     unamedkr

     and

     claude

     authored
     79db43aView commit details

     Copy full SHA for 79db43a

     View code at this pointBrowse repository at this point
7. Commits on Apr 10, 2026

   • v0.11.0: quantcpp CLI + progressive default + Llama-3.2-1B default

     Show description for bc44a6a
     

     unamedkr

     and

     claude

     committed
     bc44a6aView commit details

     Copy full SHA for bc44a6a

     View code at this pointBrowse repository at this point

   • perf(wasm): Web Worker + no ASYNCIFY — maximum inference speed (#28)

     Show description for 3f3fb74
     

     unamedkr

     and

     claude

     authored
     3f3fb74View commit details

     Copy full SHA for 3f3fb74

     View code at this pointBrowse repository at this point

   • progressive=True is now the default

     Show description for 0bc49fc
     

     unamedkr

     and

     claude

     committed
     0bc49fcView commit details

     Copy full SHA for 0bc49fc

     View code at this pointBrowse repository at this point

   • Fix Qwen RMSNorm: revert runtime +1 for GGUF + switch demo to Qwen3.5 (#24)

     Show description for 9b9ce04
     

     unamedkr

     and

     claude

     authored
     9b9ce04View commit details

     Copy full SHA for 9b9ce04

     View code at this pointBrowse repository at this point

   • feat(wasm): model selector (Qwen3 0.6B / Llama 3.2 1B) + real-time streaming (#20)

     Show description for 10c49ff
     

     unamedkr

     and

     claude

     authored
     10c49ffView commit details

     Copy full SHA for 10c49ff

     View code at this pointBrowse repository at this point
8. Commits on Apr 9, 2026

   • S1: expose aggressive=True in Python API (4-bit + k512 window)

     Show description for d07e809
     

     unamedkr

     and

     claude

     committed
     d07e809View commit details

     Copy full SHA for d07e809

     View code at this pointBrowse repository at this point

   • quantcpp 0.10.1: includes S2 infinite scrollback + S4 context persistence

     unamedkr

     committed
     cc74de2View commit details

     Copy full SHA for cc74de2

     View code at this pointBrowse repository at this point

   • S4: context persistence — save/load KV cache to disk

     Show description for 953addb
     

     unamedkr

     and

     claude

     committed
     953addbView commit details

     Copy full SHA for 953addb

     View code at this pointBrowse repository at this point

   • quantcpp 0.10.0: progressive KV compression — human-memory-like caching

     Show description for 54e8b24
     

     unamedkr

     and

     claude

     committed
     54e8b24View commit details

     Copy full SHA for 54e8b24

     View code at this pointBrowse repository at this point

   • quantcpp 0.9.2: add Llama-3.2-1B to model registry

     Show description for 6b781eb
     

     unamedkr

     and

     claude

     committed
     6b781ebView commit details

     Copy full SHA for 6b781eb

     View code at this pointBrowse repository at this point

   • quantcpp 0.9.1: context_length parameter — longer context via KV compression

     Show description for b1786b9
     

     unamedkr

     and

     claude

     committed
     b1786b9View commit details

     Copy full SHA for b1786b9

     View code at this pointBrowse repository at this point