diff --git a/html/reference-distilled.json b/html/reference-distilled.json
index d31e3fdc5..36e583ec2 100644
--- a/html/reference-distilled.json
+++ b/html/reference-distilled.json
@@ -161,5 +161,25 @@
     "preview": "Tencent-Hunyuan--HunyuanDiT-v1.1-Diffusers-Distilled.jpg",
     "tags": "distilled",
     "extras": "sampler: Default, cfg_scale: 2.0"
+  },
+  "Black Forest Labs FLUX.2 Klein 4B": {
+    "path": "black-forest-labs/FLUX.2-klein-4B",
+    "preview": "black-forest-labs--FLUX.2-klein-4B.jpg",
+    "desc": "FLUX.2-klein-4B is a 4 billion parameter size-distilled version of FLUX.2-dev optimized for consumer GPUs. Achieves sub-second inference with 4 steps while fitting in ~13GB VRAM. Supports both text-to-image generation and multi-reference image editing. Apache 2.0 licensed.",
+    "skip": true,
+    "tags": "distilled",
+    "extras": "sampler: Default, cfg_scale: 4.0, steps: 4",
+    "size": 8.5,
+    "date": "2025 January"
+  },
+  "Black Forest Labs FLUX.2 Klein 9B": {
+    "path": "black-forest-labs/FLUX.2-klein-9B",
+    "preview": "black-forest-labs--FLUX.2-klein-9B.jpg",
+    "desc": "FLUX.2-klein-9B is a 9 billion parameter size-distilled version of FLUX.2-dev. Higher quality than 4B variant with sub-second inference using 4 steps. Requires ~29GB VRAM. Supports text-to-image and multi-reference editing. Non-commercial license.",
+    "skip": true,
+    "tags": "distilled",
+    "extras": "sampler: Default, cfg_scale: 4.0, steps: 4",
+    "size": 18.5,
+    "date": "2025 January"
   }
 }
\ No newline at end of file
diff --git a/html/reference.json b/html/reference.json
index d86214467..63cd80124 100644
--- a/html/reference.json
+++ b/html/reference.json
@@ -124,6 +124,24 @@
     "size": 104.74,
     "date": "2025 November"
   },
+  "Black Forest Labs FLUX.2 Klein Base 4B": {
+    "path": "black-forest-labs/FLUX.2-klein-base-4B",
+    "preview": "black-forest-labs--FLUX.2-klein-base-4B.jpg",
+    "desc": "FLUX.2-klein-base-4B is the undistilled 4 billion parameter base model of FLUX.2-klein. Requires 50 inference steps for full quality but offers flexibility for fine-tuning. Fits in ~13GB VRAM. Supports text-to-image and multi-reference editing. Apache 2.0 licensed.",
+    "skip": true,
+    "extras": "sampler: Default, cfg_scale: 4.0, steps: 50",
+    "size": 8.5,
+    "date": "2025 January"
+  },
+  "Black Forest Labs FLUX.2 Klein Base 9B": {
+    "path": "black-forest-labs/FLUX.2-klein-base-9B",
+    "preview": "black-forest-labs--FLUX.2-klein-base-9B.jpg",
+    "desc": "FLUX.2-klein-base-9B is the undistilled 9 billion parameter base model of FLUX.2-klein. Requires 50 inference steps for full quality but offers flexibility for fine-tuning. Requires ~29GB VRAM. Supports text-to-image and multi-reference editing. Non-commercial license.",
+    "skip": true,
+    "extras": "sampler: Default, cfg_scale: 4.0, steps: 50",
+    "size": 18.5,
+    "date": "2025 January"
+  },
 
   "Z-Image-Turbo": {
     "path": "Tongyi-MAI/Z-Image-Turbo",
diff --git a/installer.py b/installer.py
index 913c7d357..08a4bc3e0 100644
--- a/installer.py
+++ b/installer.py
@@ -648,7 +648,7 @@ def check_diffusers():
     t_start = time.time()
     if args.skip_all:
         return
-    sha = '5efb81fa711863fdece9136ad10788440e658b40' # diffusers commit hash
+    sha = '61f175660a8ac54f1470a74a810e6c38fb4795d5' # diffusers commit hash
     # if args.use_rocm or args.use_zluda or args.use_directml:
     #     sha = '043ab2520f6a19fce78e6e060a68dbc947edb9f9' # lock diffusers versions for now
     pkg = pkg_resources.working_set.by_key.get('diffusers', None)
diff --git a/modules/sd_detect.py b/modules/sd_detect.py
index 0e93b6d20..a1cb6e913 100644
--- a/modules/sd_detect.py
+++ b/modules/sd_detect.py
@@ -92,6 +92,8 @@ def guess_by_name(fn, current_guess):
         new_guess = 'HiDream'
     elif 'chroma' in fn.lower() and 'xl' not in fn.lower():
         new_guess = 'Chroma'
+    elif 'flux.2' in fn.lower() and 'klein' in fn.lower():
+        new_guess = 'FLUX2 Klein'
     elif 'flux.2' in fn.lower():
         new_guess = 'FLUX2'
     elif 'flux' in fn.lower() or 'flex.1' in fn.lower():
diff --git a/modules/sd_models.py b/modules/sd_models.py
index f0a832dd6..8db759e87 100644
--- a/modules/sd_models.py
+++ b/modules/sd_models.py
@@ -359,6 +359,10 @@ def load_diffuser_force(detected_model_type, checkpoint_info, diffusers_load_con
             from pipelines.model_flux2 import load_flux2
             sd_model = load_flux2(checkpoint_info, diffusers_load_config)
             allow_post_quant = False
+        elif model_type in ['FLUX2 Klein']:
+            from pipelines.model_flux2_klein import load_flux2_klein
+            sd_model = load_flux2_klein(checkpoint_info, diffusers_load_config)
+            allow_post_quant = False
         elif model_type in ['FLEX']:
             from pipelines.model_flex import load_flex
             sd_model = load_flex(checkpoint_info, diffusers_load_config)
diff --git a/modules/shared_items.py b/modules/shared_items.py
index b5df05390..3186b3d12 100644
--- a/modules/shared_items.py
+++ b/modules/shared_items.py
@@ -48,6 +48,8 @@ pipelines = {
     'Qwen': getattr(diffusers, 'QwenImagePipeline', None),
     'HunyuanImage': getattr(diffusers, 'HunyuanImagePipeline', None),
     'Z-Image': getattr(diffusers, 'ZImagePipeline', None),
+    'FLUX2': getattr(diffusers, 'Flux2Pipeline', None),
+    'FLUX2 Klein': getattr(diffusers, 'Flux2KleinPipeline', None),
     'LongCat': getattr(diffusers, 'LongCatImagePipeline', None),
     'GLM-Image': getattr(diffusers, 'GlmImagePipeline', None),
     # dynamically imported and redefined later
diff --git a/pipelines/generic.py b/pipelines/generic.py
index d6dad46c8..0a51c1cec 100644
--- a/pipelines/generic.py
+++ b/pipelines/generic.py
@@ -200,6 +200,23 @@ def load_text_encoder(repo_id, cls_name, load_config=None, subfolder="text_encod
                 **load_args,
                 **quant_args,
             )
+        # Qwen3ForCausalLM - shared text encoders by hidden_size:
+        # - Z-Image, Klein-4B: Qwen3-4B (hidden_size=2560)
+        # - Klein-9B: Qwen3-8B (hidden_size=4096)
+        elif cls_name == transformers.Qwen3ForCausalLM and allow_shared and shared.opts.te_shared_t5:
+            if '-9b' in repo_id.lower():
+                shared_repo = 'black-forest-labs/FLUX.2-klein-9B'  # 9B variants use Qwen3-8B
+            else:
+                shared_repo = 'Tongyi-MAI/Z-Image-Turbo'  # 4B variants and Z-Image use Qwen3-4B
+            subfolder = 'text_encoder'
+            shared.log.debug(f'Load model: text_encoder="{shared_repo}" cls={cls_name.__name__} quant="{quant_type}" loader={_loader("transformers")} shared={shared.opts.te_shared_t5}')
+            text_encoder = cls_name.from_pretrained(
+                shared_repo,
+                cache_dir=shared.opts.hfcache_dir,
+                subfolder=subfolder,
+                **load_args,
+                **quant_args,
+            )
 
         # load from repo
         if text_encoder is None:
diff --git a/pipelines/model_flux2_klein.py b/pipelines/model_flux2_klein.py
new file mode 100644
index 000000000..9b8b05ac8
--- /dev/null
+++ b/pipelines/model_flux2_klein.py
@@ -0,0 +1,42 @@
+import transformers
+import diffusers
+from modules import shared, devices, sd_models, model_quant, sd_hijack_te, sd_hijack_vae
+from pipelines import generic
+
+
+def load_flux2_klein(checkpoint_info, diffusers_load_config=None):
+    if diffusers_load_config is None:
+        diffusers_load_config = {}
+    repo_id = sd_models.path_to_repo(checkpoint_info)
+    sd_models.hf_auth_check(checkpoint_info)
+
+    load_args, _quant_args = model_quant.get_dit_args(diffusers_load_config, allow_quant=False)
+    shared.log.debug(f'Load model: type=Flux2Klein repo="{repo_id}" config={diffusers_load_config} offload={shared.opts.diffusers_offload_mode} dtype={devices.dtype} args={load_args}')
+
+    # Load transformer - Klein uses Flux2Transformer2DModel (same class as Flux2, different size)
+    transformer = generic.load_transformer(repo_id, cls_name=diffusers.Flux2Transformer2DModel, load_config=diffusers_load_config)
+
+    # Load text encoder - Klein uses Qwen3ForCausalLM (8B), shared across all Klein variants
+    text_encoder = generic.load_text_encoder(repo_id, cls_name=transformers.Qwen3ForCausalLM, load_config=diffusers_load_config)
+
+    pipe = diffusers.Flux2KleinPipeline.from_pretrained(
+        repo_id,
+        transformer=transformer,
+        text_encoder=text_encoder,
+        cache_dir=shared.opts.diffusers_dir,
+        **load_args,
+    )
+    pipe.task_args = {
+        'output_type': 'np',
+    }
+    diffusers.pipelines.auto_pipeline.AUTO_TEXT2IMAGE_PIPELINES_MAPPING["flux2klein"] = diffusers.Flux2KleinPipeline
+    diffusers.pipelines.auto_pipeline.AUTO_IMAGE2IMAGE_PIPELINES_MAPPING["flux2klein"] = diffusers.Flux2KleinPipeline
+    diffusers.pipelines.auto_pipeline.AUTO_INPAINT_PIPELINES_MAPPING["flux2klein"] = diffusers.Flux2KleinPipeline
+
+    del text_encoder
+    del transformer
+    sd_hijack_te.init_hijack(pipe)
+    sd_hijack_vae.init_hijack(pipe)
+
+    devices.torch_gc(force=True, reason='load')
+    return pipe