fix: use correct model names, async batch examples, and cost savings

- Replace gpt-4o with Qwen/Qwen3-30B-A3B and text-embedding-3-small
  with Qwen/Qwen3-Embedding-8B (models available through Doubleword)
- Fix batch class examples to use async methods (ainvoke, aembed_documents)
  since ChatDoublewordBatch and DoublewordEmbeddingsBatch are async-only
- Update cost savings messaging to "up to 90% cost savings"

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: sejori

src/oss/python/integrations/chat/doubleword.mdx

@@ -71,7 +71,7 @@ Now we can instantiate our model object and generate chat completions:
 from langchain_doubleword import ChatDoubleword
 
 model = ChatDoubleword(
-    model="gpt-4o",
+    model="Qwen/Qwen3-30B-A3B",
     temperature=0,
     max_tokens=1024,
     max_retries=2,
@@ -142,19 +142,29 @@ For more on binding tools and tool call outputs, head to the [tool calling](/oss
 
 ## Batch processing
 
-`ChatDoublewordBatch` uses Doubleword's batch API to transparently collect concurrent calls into batch submissions at reduced cost. This is useful for high-throughput workloads where real-time responses are not required.
+`ChatDoublewordBatch` uses Doubleword's batch API to transparently collect concurrent calls into batch submissions with up to 90% cost savings. This is useful for high-throughput workloads where real-time responses are not required.
+
+**Note:** `ChatDoublewordBatch` is async-only. Sync methods like `invoke()` will raise `NotImplementedError`. Use `ainvoke()` instead.
 
 ```python
+import asyncio
 from langchain_doubleword import ChatDoublewordBatch
 
 batch_model = ChatDoublewordBatch(
-    model="gpt-4o",
+    model="Qwen/Qwen3-30B-A3B",
     temperature=0,
 )
 
+
 # Calls are automatically batched behind the scenes
-result = batch_model.invoke("Summarize the theory of relativity in one sentence.")
-result.content
+async def main():
+    result = await batch_model.ainvoke(
+        "Summarize the theory of relativity in one sentence."
+    )
+    print(result.content)
+
+
+asyncio.run(main())
 ```
 
 ---

src/oss/python/integrations/embeddings/doubleword.mdx

@@ -34,7 +34,7 @@ if not os.getenv("DOUBLEWORD_API_KEY"):
 ```python
 from langchain_doubleword import DoublewordEmbeddings
 
-embeddings = DoublewordEmbeddings(model="text-embedding-3-small")
+embeddings = DoublewordEmbeddings(model="Qwen/Qwen3-Embedding-8B")
 
 # Embed a single query
 query_embedding = embeddings.embed_query("What is the meaning of life?")
@@ -47,15 +47,25 @@ doc_embeddings = embeddings.embed_documents(
 
 ## Batch embeddings
 
-For high-throughput workloads, use `DoublewordEmbeddingsBatch` to automatically batch concurrent embedding requests at reduced cost:
+For high-throughput workloads, use `DoublewordEmbeddingsBatch` to automatically batch concurrent embedding requests with up to 90% cost savings.
+
+**Note:** `DoublewordEmbeddingsBatch` is async-only. Sync methods like `embed_documents()` will raise `NotImplementedError`. Use `aembed_documents()` instead.
 
 ```python
+import asyncio
 from langchain_doubleword import DoublewordEmbeddingsBatch
 
-batch_embeddings = DoublewordEmbeddingsBatch(model="text-embedding-3-small")
-doc_embeddings = batch_embeddings.embed_documents(
-    ["Document one.", "Document two.", "Document three."]
-)
+batch_embeddings = DoublewordEmbeddingsBatch(model="Qwen/Qwen3-Embedding-8B")
+
+
+async def main():
+    doc_embeddings = await batch_embeddings.aembed_documents(
+        ["Document one.", "Document two.", "Document three."]
+    )
+    print(f"Generated {len(doc_embeddings)} embeddings")
+
+
+asyncio.run(main())
 ```
 
 ## API reference

src/oss/python/integrations/providers/doubleword.mdx

@@ -4,7 +4,7 @@ description: "Route AI inference through Doubleword's unified gateway using Lang
 sidebarTitle: "Doubleword"
 ---
 
-[Doubleword](https://doubleword.ai/) is an AI model gateway and control layer that provides unified routing, management, and security for inference across multiple model providers. It exposes an OpenAI-compatible API with features like per-key rate limiting, request logging, and cost-optimized batch processing.
+[Doubleword](https://doubleword.ai/) is an AI model gateway and control layer that provides unified routing, management, and security for inference across multiple model providers. It exposes an OpenAI-compatible API with features like per-key rate limiting, request logging, and cost-optimized batch processing with up to 90% cost savings.
 
 ## Chat models