From dd75e615883cb7d84e60dd42f07b37c22cd525be Mon Sep 17 00:00:00 2001
From: Prabod Rathnayaka <praboderathnayaka@gmail.com>
Date: Fri, 31 Jul 2026 21:00:39 +1000
Subject: [PATCH] docs(changelog): note the CUDA pull-variant fix re-uploaded
 into v0.2.0

---
 CHANGELOG.md | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/CHANGELOG.md b/CHANGELOG.md
index aa53e38..110e4f5 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -37,6 +37,10 @@ overhaul of `basert serve`.
 
 ### Fixed
 
+- **CUDA model resolution.** `basert pull`/`chat <id>` on NVIDIA GB10 now fetches
+  the CUDA-native bundle instead of the universal one it couldn't load
+  (`no function named 'gemv_q4'`). The 0.2.0 Linux/arm64 CUDA download was updated
+  in place with this fix.
 - **Chunked-decode correctness on CUDA.** Free generation past the decode chunk size
   no longer repeats an earlier block (an eager-dispatch write-after-read hazard);
   prefill and teacher-forced decode were always correct.