diff --git a/README.md b/README.md
index f016288..23a7668 100644
--- a/README.md
+++ b/README.md
@@ -29,7 +29,7 @@ Add `:emily` to your `mix.exs` deps:
 ```elixir
 def deps do
   [
-    {:emily, "~> 0.7"}
+    {:emily, "~> 1.0"}
   ]
 end
 ```
@@ -65,9 +65,11 @@ Emily backend smoke test.
   other backends.
 - **Affine group-wise quantization.** `Emily.QuantizedWeight` +
   `Emily.Quantization` wrap MLX `quantize` / `dequantize` /
-  `quantized_matmul` for int2 / int4 / int8 inference. Includes a
-  defn-native `dequantize_defn/1` for quantized layers inside Axon
-  forward passes.
+  `quantized_matmul` for int2 / int4 / int8 inference. Quantized dense
+  layers lower to the fused `quantized_matmul` kernel — streaming the
+  packed low-bit weights instead of dequantizing the whole weight per
+  token — with a `dequantize_defn/1` + `Nx.dot` fallback on non-MLX
+  backends.
 - **Mixed-precision training.** `Emily.MixedPrecision` provides the
   bf16 recipe (cast params for the forward, keep f32 master, dynamic
   loss scaling with overflow detection).
@@ -99,7 +101,7 @@ Releases on first `mix compile`.
 
 ### As a hex consumer
 
-Add `{:emily, "~> 0.7"}` to `mix.exs`, then:
+Add `{:emily, "~> 1.0"}` to `mix.exs`, then:
 
 ```sh
 mix deps.get
diff --git a/lib/emily/bumblebee/fast_kernels.ex b/lib/emily/bumblebee/fast_kernels.ex
index cd0ab1e..729f5f2 100644
--- a/lib/emily/bumblebee/fast_kernels.ex
+++ b/lib/emily/bumblebee/fast_kernels.ex
@@ -24,8 +24,8 @@ if Code.ensure_loaded?(Axon) and Code.ensure_loaded?(Bumblebee.Layers) do
     entirely when either dep is missing. To use the shim, add both to
     your own `deps/0`:
 
-        {:bumblebee, "~> 0.6"},
-        {:axon, "~> 0.7"}
+        {:bumblebee, "~> 0.7"},
+        {:axon, "~> 0.8"}
 
     ## Usage
 
diff --git a/livebooks/distilbert_qa.livemd b/livebooks/distilbert_qa.livemd
index 19cabf7..62d77af 100644
--- a/livebooks/distilbert_qa.livemd
+++ b/livebooks/distilbert_qa.livemd
@@ -3,7 +3,7 @@
 ```elixir
 Mix.install(
   [
-    {:emily, "~> 0.7"},
+    {:emily, "~> 1.0"},
     {:bumblebee, "~> 0.7"},
     {:tokenizers, "~> 0.5"},
     {:nx, "~> 0.12"},
diff --git a/livebooks/fast_kernels.livemd b/livebooks/fast_kernels.livemd
index e121b7f..16c4bb0 100644
--- a/livebooks/fast_kernels.livemd
+++ b/livebooks/fast_kernels.livemd
@@ -3,7 +3,7 @@
 ```elixir
 Mix.install(
   [
-    {:emily, "~> 0.7"},
+    {:emily, "~> 1.0"},
     {:bumblebee, "~> 0.7"},
     {:tokenizers, "~> 0.5"},
     {:axon, "~> 0.8"},
diff --git a/livebooks/mnist_training.livemd b/livebooks/mnist_training.livemd
index 780db83..31b9df0 100644
--- a/livebooks/mnist_training.livemd
+++ b/livebooks/mnist_training.livemd
@@ -3,7 +3,7 @@
 ```elixir
 Mix.install(
   [
-    {:emily, "~> 0.7"},
+    {:emily, "~> 1.0"},
     {:axon, "~> 0.8"},
     {:scidata, "~> 0.1"},
     {:nx, "~> 0.12"},
diff --git a/livebooks/modernbert_classification.livemd b/livebooks/modernbert_classification.livemd
index 99fe637..53b7bdb 100644
--- a/livebooks/modernbert_classification.livemd
+++ b/livebooks/modernbert_classification.livemd
@@ -3,7 +3,7 @@
 ```elixir
 Mix.install(
   [
-    {:emily, "~> 0.7"},
+    {:emily, "~> 1.0"},
     {:bumblebee, "~> 0.7"},
     {:tokenizers, "~> 0.5"},
     {:nx, "~> 0.12"},
diff --git a/livebooks/nomic_embeddings.livemd b/livebooks/nomic_embeddings.livemd
index 4ca194a..434709d 100644
--- a/livebooks/nomic_embeddings.livemd
+++ b/livebooks/nomic_embeddings.livemd
@@ -3,7 +3,7 @@
 ```elixir
 Mix.install(
   [
-    {:emily, "~> 0.7"},
+    {:emily, "~> 1.0"},
     {:bumblebee, "~> 0.7"},
     {:tokenizers, "~> 0.5"},
     {:nx, "~> 0.12"},
diff --git a/livebooks/qwen3_quantized.livemd b/livebooks/qwen3_quantized.livemd
index 128501a..7fba122 100644
--- a/livebooks/qwen3_quantized.livemd
+++ b/livebooks/qwen3_quantized.livemd
@@ -3,7 +3,7 @@
 ```elixir
 Mix.install(
   [
-    {:emily, "~> 0.7"},
+    {:emily, "~> 1.0"},
     {:bumblebee, "~> 0.7"},
     {:tokenizers, "~> 0.5"},
     {:nx, "~> 0.12"},
@@ -57,7 +57,7 @@ dense_text
 `native: true, native_fallback: :raise` on `:defn_options` lowers the
 per-token decode through Emily's native Expr compiler and asserts it
 stays fully native — the quantized serving below passes the same
-options, so the int4 dequantize path lowers native too.
+options, so the int4 quantized matmul lowers native too.
 
 ## Quantization transform
 
@@ -168,8 +168,8 @@ end
 > recursive rewriter. The `transpose: true` default stores weights as
 > `[out, in]` (MLX / AWQ convention, groups along the reduction axis);
 > set `false` if you're feeding a checkpoint that's already laid out
-> the other way. Bits must be one of `[2, 3, 4, 6, 8]` for the
-> defn-native `dequantize_defn/1` path.
+> the other way. Bits must be one of `[2, 3, 4, 6, 8]` for MLX affine
+> quantization.
 
 ## Quantized inference
 
diff --git a/livebooks/smollm3_chat.livemd b/livebooks/smollm3_chat.livemd
index ae595a1..a4e731b 100644
--- a/livebooks/smollm3_chat.livemd
+++ b/livebooks/smollm3_chat.livemd
@@ -3,7 +3,7 @@
 ```elixir
 Mix.install(
   [
-    {:emily, "~> 0.7"},
+    {:emily, "~> 1.0"},
     {:bumblebee, "~> 0.7"},
     {:tokenizers, "~> 0.5"},
     {:nx, "~> 0.12"},
diff --git a/livebooks/whisper_transcription.livemd b/livebooks/whisper_transcription.livemd
index be2f14f..efb42d1 100644
--- a/livebooks/whisper_transcription.livemd
+++ b/livebooks/whisper_transcription.livemd
@@ -3,7 +3,7 @@
 ```elixir
 Mix.install(
   [
-    {:emily, "~> 0.7"},
+    {:emily, "~> 1.0"},
     {:bumblebee, "~> 0.7"},
     {:tokenizers, "~> 0.5"},
     {:nx, "~> 0.12"},