Release 0.5.0

CHANGELOG.md

@@ -2,6 +2,33 @@
 
 All notable changes to GPULlama3.java will be documented in this file.
 
+## [0.5.0] - 2026-06-24
+
+### Features
+
+- Add prefill-decode and batch-prefill-decode for Qwen3 (FP16 and Q8_0) ([#122](https://github.com/beehive-lab/GPULlama3.java/pull/122))
+- Refactor GPU backend planner ([#117](https://github.com/beehive-lab/GPULlama3.java/pull/117))
+- Several fixes and improvements for CI ([#115](https://github.com/beehive-lab/GPULlama3.java/pull/115))
+- Ci/metrics history ([#114](https://github.com/beehive-lab/GPULlama3.java/pull/114))
+- Improve collection of performance/throughput metrics ([#113](https://github.com/beehive-lab/GPULlama3.java/pull/113))
+- Update TornadoVM dependency for jdk21 and fixed suffix regarding future releases ([#111](https://github.com/beehive-lab/GPULlama3.java/pull/111))
+- Add Prefill–Decode Separation with Batched Prompt Ingestion and Logits Skipping  ([#102](https://github.com/beehive-lab/GPULlama3.java/pull/102))
+
+### Other Changes
+
+- Qwen3 decode: split-KV attention + backend-aware warp GEMV (FP16 & Q8_0) ([#123](https://github.com/beehive-lab/GPULlama3.java/pull/123))
+- Introduce tool calling support ([#116](https://github.com/beehive-lab/GPULlama3.java/pull/116))
+- Cleanup of presentation materials ([#121](https://github.com/beehive-lab/GPULlama3.java/pull/121))
+- Add Q4_K/Q5_K/Q6_K GPU support via Q8_0 dequantization ([#108](https://github.com/beehive-lab/GPULlama3.java/pull/108))
+- llama-tornado script curation ([#112](https://github.com/beehive-lab/GPULlama3.java/pull/112))
+- Add Apple Metal backend support ([#103](https://github.com/beehive-lab/GPULlama3.java/pull/103))
+- Add DevoxxGreece presentation material ([#109](https://github.com/beehive-lab/GPULlama3.java/pull/109))
+-   Devstral 2 support (Mistral 3 architecture, Tekken tokenizer, YaRN … ([#107](https://github.com/beehive-lab/GPULlama3.java/pull/107))
+-   Add llamaTornado Java 25 single-file launcher with Metal backend support   ([#105](https://github.com/beehive-lab/GPULlama3.java/pull/105))
+- [refactor] Simplify and unify the TornadoVM layer planner infrastructure ([#101](https://github.com/beehive-lab/GPULlama3.java/pull/101))
+- AddCI Actions for Quarkus-LangChain4j integration ([#89](https://github.com/beehive-lab/GPULlama3.java/pull/89))
+- Simplify and generalize TornadoVM version across JDK profiles in pom.xml ([#99](https://github.com/beehive-lab/GPULlama3.java/pull/99))
+
 ## [0.4.0] - 2026-02-25
 
 ### Other Changes

CITATION.cff

@@ -15,6 +15,6 @@ authors:
   given-names: "Christos"
 title: "GPULlama3.java"
 license: MIT License
-version: 0.4.0
-date-released: 2026-02-25
+version: 0.5.0
+date-released: 2026-06-24
 url: "https://github.com/beehive-lab/GPULlama3.java"

README.md

@@ -133,7 +133,7 @@ You can add **GPULlama3.java** directly to your Maven project by including the f
 <dependency>
     <groupId>io.github.beehive-lab</groupId>
     <artifactId>gpu-llama3</artifactId>
-    <version>0.4.0</version>
+    <version>0.5.0</version>
 </dependency>
 ```
 
@@ -142,7 +142,7 @@ You can add **GPULlama3.java** directly to your Maven project by including the f
 <dependency>
     <groupId>io.github.beehive-lab</groupId>
     <artifactId>gpu-llama3</artifactId>
-    <version>0.4.0-jdk25</version>
+    <version>0.5.0-jdk25</version>
 </dependency>
 ```
 

pom.xml

@@ -38,7 +38,7 @@
 
         <properties>
             <!-- CI-friendly version: resolved by flatten-maven-plugin at build time -->
-            <revision>0.4.0</revision>
+            <revision>0.5.0</revision>
             <tornadovm.base.version>4.0.2</tornadovm.base.version>
             <jdk.version.suffix>-jdk21</jdk.version.suffix>
             <tornadovm.version>${tornadovm.base.version}${jdk.version.suffix}</tornadovm.version>