Qwen/Qwen3-14B
Run locally on Apple devices with Mirai
- Type
- local
- From
- Alibaba
- Quantization
- No
- Parameters
- 14B
- Size
- 27.5 GB
- Source
Qwen3-14B is a 14.8-billion parameter causal language model from Alibaba's Qwen team, representing the latest generation in the Qwen series. It introduces a distinctive dual-mode architecture that allows seamless switching between a thinking mode — designed for complex reasoning tasks like math, coding, and logic — and a non-thinking mode for efficient, general-purpose conversation.
Architecture & Specifications
- Parameters: 14.8B total (13.2B non-embedding)
- Layers: 40, with Grouped Query Attention (40 Q heads, 8 KV heads)
- Context Length: 32,768 tokens natively; up to 131,072 tokens via YaRN RoPE scaling
- Base Model: Qwen3-14B-Base, with both pretraining and post-training stages
- License: Apache 2.0
Key Capabilities
Dual Reasoning Modes. The model's thinking mode wraps internal chain-of-thought reasoning inside `<think>...</think>` blocks before delivering a final answer, surpassing previous QwQ models on mathematical and code-generation benchmarks. Non-thinking mode mirrors the behavior of Qwen2.5-Instruct for fast, direct responses. Users can toggle modes via a simple `enable_thinking` flag or inline `/think` and `/no_think` commands within conversation turns.
Agentic & Tool Use. Qwen3-14B excels at structured tool calling and integration with external services, including MCP-based tool configurations. It achieves leading performance among open-source models on complex agent-based tasks.
Multilingual Breadth. The model supports over 100 languages and dialects, with strong multilingual instruction-following and translation capabilities.
Human Preference Alignment. Post-training emphasizes creative writing, role-playing, multi-turn dialogue, and precise instruction following, resulting in a natural and engaging conversational style.
Ideal Use Cases
- Complex reasoning, math problem-solving, and code generation (thinking mode)
- Conversational assistants and chatbots requiring low latency (non-thinking mode)
- Agentic workflows with external tool integration
- Multilingual applications and translation pipelines
Qwen3-14B is compatible with Hugging Face Transformers, vLLM, SGLang, Ollama, llama.cpp, and other major inference frameworks.
spm https://github.com/trymirai/uzu.git
| 1 | import Uzu |
| 2 | |
| 3 | public func runChat() async throws { |
| 4 | let engineConfig = EngineConfig.create() |
| 5 | let engine = try await Engine.create(config: engineConfig) |
| 6 | |
| 7 | guard let model = try await engine.model(identifier: "Qwen/Qwen3-14B") else { |
| 8 | return |
| 9 | } |
| 10 | for try await update in try await engine.download(model: model).iterator() { |
| 11 | print("Download progress: \(update.progress())") |
| 12 | } |
| 13 | |
| 14 | let messages = [ |
| 15 | ChatMessage.system().withText(text: "You are a helpful assistant"), |
| 16 | ChatMessage.user().withText(text: "Tell me a short, funny story about a robot") |
| 17 | ] |
| 18 | let session = try await engine.chat(model: model, config: .create()) |
| 19 | let stream = await session.replyWithStream(input: messages, config: .create()) |
| 20 | var message: ChatMessage? = nil |
| 21 | for try await update in stream.iterator() { |
| 22 | switch update { |
| 23 | case .replies(let replies): |
| 24 | message = replies.last?.message |
| 25 | case .error(let error): |
| 26 | print("Error: \(error)") |
| 27 | } |
| 28 | } |
| 29 | print("Text: \(message?.text() ?? "empty")") |
| 30 | } |