diff --git a/CHANGELOG.md b/CHANGELOG.md
index d425cb08..9c7ae81a 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -5,6 +5,51 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [Unreleased](https://github.com/elixir-nx/bumblebee)
+
+This release changes the directory structure of the models cache, such that cached files from the same HuggingFace Hub repository are grouped in a separate subdirectory. This change is meant to simplify the process of manually removing specific models from the cache to free up space. As a result, the cache contents from prior versions are invalidated, so you most likely want to remove the current cache contents. To find the cache location run `elixir -e 'Mix.install([{:bumblebee, "0.4.2"}]); IO.puts(Bumblebee.cache_dir())'` (defaults to the standard cache location for the given operating system).
+
+### Added
+
+* Notebook on LLaMA 2 to the docs ([#259](https://github.com/elixir-nx/bumblebee/pull/259))
+* Mistral model ([#264](https://github.com/elixir-nx/bumblebee/pull/264))
+* Projection head models for ClipText and ClipVision ([#276](https://github.com/elixir-nx/bumblebee/pull/276))
+* Support more rotary embedding options for LLaMA required for Deepseek Coder ([#285](https://github.com/elixir-nx/bumblebee/pull/285))
+* Temperature generation option ([#290](https://github.com/elixir-nx/bumblebee/pull/290))
+* GPTBigCode model (used by Starcoder) ([#294](https://github.com/elixir-nx/bumblebee/pull/294))
+* Automatic detection of diffusers params files (specifying `:params_filename` for Stable Diffusion models is no longer necessary) ([#301](https://github.com/elixir-nx/bumblebee/pull/301))
+* `:seed` option to generation serving inputs ([#303](https://github.com/elixir-nx/bumblebee/pull/303))
+* `:params_variant` option to `Bumblebee.load_model/2` for loading parameters of different precision ([#309](https://github.com/elixir-nx/bumblebee/pull/309))
+* `:type` option to `Bumblebee.load_model/2` for loading model under a specific precision policy ([#311](https://github.com/elixir-nx/bumblebee/pull/311))
+* LCM scheduler ([#320](https://github.com/elixir-nx/bumblebee/pull/320))
+
+### Changed
+
+* **(Breaking)** Text generation to always return only the new text (for some models it used to include the prompt) ([#302](https://github.com/elixir-nx/bumblebee/pull/302))
+* Deprecated all options in `Bumblebee.apply_tokenizer/3`, these should now be set on the tokenizer using `Bumblebee.configure/2` ([#310](https://github.com/elixir-nx/bumblebee/pull/310))
+* Reduced memory used when the `:preallocate_params` serving option is enabled ([#317](https://github.com/elixir-nx/bumblebee/pull/317))
+* **(Breaking)** Changed image size to maps in image featurizers ([#329](https://github.com/elixir-nx/bumblebee/pull/329))
+* **(Breaking)** Renamed ViT and DeiT `:for_masked_image_modeling` output from `:logits` to `:pixel_values`
+* **(Breaking)** Renamed CLIP outputs `:text_embeddings` and `:image_embeddings` to singular
+* **(Breaking)** Changed ResNet `:pooled_state` output to flatten the extra 1-sized axes
+* Cache directory structure to group files by repository ([#332](https://github.com/elixir-nx/bumblebee/pull/332))
+
+### Removed
+
+* Removed the serving `:seed` option in favour of a runtime, per-input seed ([#303](https://github.com/elixir-nx/bumblebee/pull/303))
+* Conversational serving ([#308](https://github.com/elixir-nx/bumblebee/pull/308))
+* Specific tokenizer modules in favour of a single module ([#310](https://github.com/elixir-nx/bumblebee/pull/310))
+
+### Fixed
+
+* Featurizer batch template when image size is a tuple
+* Error in concatenating results when running servings as partitioned ([#282](https://github.com/elixir-nx/bumblebee/pull/282))
+* Decoder cache being casted with low precision policies ([#299](https://github.com/elixir-nx/bumblebee/pull/299))
+* Loading of more recent VAE KL checkpoints ([#305](https://github.com/elixir-nx/bumblebee/pull/305))
+* Tokenizers truncation to account for trailing special tokens ([#307](https://github.com/elixir-nx/bumblebee/pull/307))
+* Loading models with auth token from within a HuggingFace Space ([#314](https://github.com/elixir-nx/bumblebee/pull/314))
+* Zero-shot classification serving to handle uppercased entailment token in model config ([#327](https://github.com/elixir-nx/bumblebee/pull/327))
+
 ## [v0.4.2](https://github.com/elixir-nx/bumblebee/tree/v0.4.2) (2023-09-28)
 
 ### Added