diff --git a/.claude/CLAUDE.md b/.claude/CLAUDE.md
index f12b07a..f0cae2c 100644
--- a/.claude/CLAUDE.md
+++ b/.claude/CLAUDE.md
@@ -1,12 +1,12 @@
 # Soup CLI — Project CLAUDE.md
 
-Soup is a CLI-first LLM fine-tuning tool (v0.13.0). Python 3.9+, MIT license.
+Soup is a CLI-first LLM fine-tuning tool (v0.13.2). Python 3.9+, MIT license.
 
 ## Build & Development
 
 ```bash
 pip install -e ".[dev]"          # Install editable + test deps
-pytest tests/ -v --tb=short      # Run all tests (906 tests)
+pytest tests/ -v --tb=short      # Run all tests (917 tests)
 ruff check soup_cli/ tests/      # Lint (must pass before commit)
 ruff check --fix soup_cli/ tests/  # Auto-fix lint issues
 ```
@@ -70,7 +70,7 @@ soup_cli/
     vllm.py            # AsyncLLMEngine backend (2-4x inference throughput)
     galore.py          # GaLore optimizer config + validation
     constants.py       # APP_NAME, paths, default chat template
-tests/                 # 44 test files, 906 tests
+tests/                 # 44 test files, 917 tests
 examples/
   configs/             # 7 production-ready YAML examples
   data/                # Sample datasets
@@ -155,8 +155,9 @@ soup version           # Show version (--full for details)
 - **Deprecated CLI secrets**: `--api-key` and `--token` flags read from env vars, marked deprecated
 - **Custom reward warning**: Prominent warning before executing arbitrary .py reward files
 - **max_tokens bound**: Capped at 16384 on inference endpoints
-- **experiment_name validation**: Path separators and null bytes blocked (v0.13.0)
-- **GaLore params**: Type-enforced before string interpolation (v0.13.0)
+- **experiment_name validation**: Path separators and null bytes blocked (v0.12.0)
+- **GaLore params**: Type-enforced before string interpolation (v0.12.0)
+- **Batch inference**: max_tokens capped at 16384, trust_remote_code warning (v0.13.0)
 
 ## Code Conventions
 
@@ -217,7 +218,7 @@ soup version           # Show version (--full for details)
 12. **Tag**: `git tag v0.X.Y && git push origin v0.X.Y`
 13. **Release**: `gh release create v0.X.Y` with changelog (What's New, Install/Upgrade)
 
-## Tests (44 test files, 906 tests)
+## Tests (44 test files, 917 tests)
 
 | File | Covers |
 |------|--------|
diff --git a/README.md b/README.md
index 8a626dd..7d7a65e 100644
--- a/README.md
+++ b/README.md
@@ -783,7 +783,7 @@ soup version
 
 # Full system info (useful for bug reports)
 soup version --full
-# -> soup v0.13.0 | Python 3.11.5 | CUDA 12.1 | extras: serve, data
+# -> soup v0.13.2 | Python 3.11.5 | CUDA 12.1 | extras: serve, data
 ```
 
 ## Web UI
@@ -953,20 +953,47 @@ soup --verbose <command>                      Full traceback on errors
 
 ## Supported Models
 
-Soup works with any HuggingFace-compatible causal language model. Tested and recommended:
+Soup works with **any** of the **340,000+** text-generation models on [HuggingFace Hub](https://huggingface.co/models?pipeline_tag=text-generation). If a model supports `AutoModelForCausalLM`, it works with Soup — zero config changes needed.
 
-| Model Family | Example Models | Notes |
+### Recommended Models
+
+| Model Family | Models | Sizes | Best For |
+|---|---|---|---|
+| **Llama 4** | Llama-4-Scout-17B, Llama-4-Maverick-17B | 17B | General, multilingual |
+| **Llama 3.x** | Llama-3.1-8B-Instruct, Llama-3.3-70B-Instruct | 1B–70B | Chat, instruction following |
+| **Llama 3.2 Vision** | Llama-3.2-11B-Vision-Instruct, Llama-3.2-90B-Vision | 11B–90B | Image understanding |
+| **Gemma 3** | Gemma-3-4B-IT, Gemma-3-9B-IT, Gemma-3-27B-IT | 4B–27B | Efficient, multilingual |
+| **Qwen 3** | Qwen3-8B, Qwen3-14B, Qwen3-32B, Qwen3-235B-A22B | 0.6B–235B | Reasoning, code, MoE |
+| **Qwen 2.5** | Qwen2.5-7B-Instruct, Qwen2.5-Coder-32B-Instruct | 0.5B–72B | Code, math |
+| **DeepSeek** | DeepSeek-R1-Distill-Llama-8B, DeepSeek-V3-0324 | 1.5B–671B | Reasoning (GRPO), code |
+| **Phi-4** | Phi-4-14B, Phi-4-mini-reasoning | 3.8B–14B | Compact reasoning |
+| **Mistral** | Mistral-7B-Instruct-v0.3, Mistral-Small-24B-Instruct | 7B–24B | Fast, efficient |
+| **Mixtral** | Mixtral-8x7B-Instruct-v0.1, Mixtral-8x22B | 47B–141B | MoE architecture |
+| **CodeLlama** | CodeLlama-7b-Instruct-hf, CodeLlama-34b-Instruct | 7B–34B | Code generation |
+| **StarCoder 2** | StarCoder2-15B, StarCoder2-7B | 3B–15B | Code completion |
+| **Yi** | Yi-1.5-34B-Chat, Yi-1.5-9B-Chat | 6B–34B | Multilingual chat |
+| **InternLM 3** | InternLM3-8B-Instruct | 8B | Chinese + English |
+| **Falcon** | Falcon-11B, Falcon-40B-Instruct | 7B–180B | Open-weight |
+
+### Vision Models (with `modality: vision`)
+
+| Model | Size | Supported Formats |
 |---|---|---|
-| **Llama 4** | Llama-4-Scout-17B, Llama-4-Maverick-17B | Latest Meta models |
-| **Llama 3.x** | Llama-3.1-8B-Instruct, Llama-3.2-11B-Vision | Text + vision |
-| **Gemma 3** | Gemma-3-9B-IT, Gemma-3-27B-IT | Google's latest |
-| **Qwen 2.5/3** | Qwen2.5-7B-Instruct, Qwen3-8B | Alibaba series |
-| **Phi-4** | Phi-4-14B | Microsoft |
-| **DeepSeek** | DeepSeek-R1-Distill-Llama-8B, DeepSeek-V3 | Reasoning models (GRPO) |
-| **Mistral** | Mistral-7B-Instruct-v0.3, Mixtral-8x7B | Including MoE |
-| **CodeLlama** | CodeLlama-7b-Instruct-hf | Code generation |
+| LLaMA-3.2-11B-Vision-Instruct | 11B | LLaVA, ShareGPT4V |
+| Qwen2-VL-7B-Instruct | 7B | LLaVA, ShareGPT4V |
+| Pixtral-12B-2409 | 12B | LLaVA, ShareGPT4V |
 
-Any model on HuggingFace Hub that supports `AutoModelForCausalLM` will work. For vision tasks, use models with vision capabilities (LLaMA-3.2-Vision, Qwen2-VL, Pixtral).
+### Quick Size Guide
+
+| VRAM | Max Model (QLoRA 4-bit) | Example |
+|---|---|---|
+| 8 GB | ~7B | Llama-3.1-8B, Mistral-7B |
+| 16 GB | ~14B | Phi-4-14B, Qwen2.5-14B |
+| 24 GB | ~34B | CodeLlama-34B, Yi-1.5-34B |
+| 48 GB | ~70B | Llama-3.3-70B |
+| 80 GB+ | 70B+ (full) or MoE | Mixtral-8x22B, DeepSeek-V3 |
+
+> **Note:** Soup auto-detects your GPU and estimates the optimal batch size. Use `soup doctor` to check your setup.
 
 ## Requirements