Thinking-mode switching
According to the model card, one model covers complex reasoning and efficient dialogue via an enable_thinking switch plus /think and /no_think prompt controls.
Open Source Model Profile · Qwen
Qwen3-1.7B-MLX-bf16
Qwen3-1.7B-MLX-bf16 is a 1.72B-parameter Qwen3 MLX build from Qwen. According to the model card, it switches between thinking mode for reasoning tasks and non-thinking mode for efficient dialogue.
- Publisher
- Qwen
- Task
- text-generation
- Model type
- qwen3
- License
- apache-2.0
- Library
- mlx
- Publication status
- Approved for indexing
Model overview
Qwen3-1.7B-MLX-bf16 is published by Qwen as an MLX text-generation build of Qwen3-1.7B. The captured configuration identifies Qwen3ForCausalLM with model type qwen3. According to the model card, Qwen3 supports switching between a thinking mode for reasoning, math, and coding and a non-thinking mode for efficient dialogue.
Recorded capabilities
Documented agent integration
According to the model card, the model integrates with external tools in both modes through SGLang, vLLM, and qwen-agent recipes.
MLX load-and-generate recipe
According to the model card, the MLX build loads with a chat template and generates responses with configurable maximum tokens.
Task-specific best practices
According to the model card, guidance covers thinking-mode sampling, boxed math answers, structured multiple-choice fields, and excluding thinking content from history.
Use cases in the source record
- Reasoning-heavy chats for math and coding run in thinking mode with turn-level /think control.
- Agent workflows that connect the model to external tools through OpenAI-compatible servers or qwen-agent.
- Efficient general dialogue and structured question answering in non-thinking mode with boxed or JSON-style answers.
Limitations and unknowns
- No evaluation results were extracted from this record.
- Layer counts, head counts, and context length are publisher model-card claims and were not independently verified.
- Provider state is historical snapshot data, not independently refreshed current availability.
Source and provenance
Source: Qwen/Qwen3-1.7B-MLX-bf16
Captured: Unknown. Processed: 2026-09-07T19:35:51.234180+00:00.
Qwen3-1.7B-MLX-bf16 Qwen3 Highlights Qwen3 is the latest generation of large language models in Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models. Built upon extensive training, Qwen3 delivers groundbreaking advancements in reasoning, instruction-following, agent capabilities, and multilingual support, with the following key features: Uniquely support of seamless switching between thinking mode (for complex logical reasoning, math, and coding) and non-thinking mode (for efficient, general-purpose dialogue) within single model , ensuring optimal performance across various scenarios. Significantl…
F001F002F003F004F005F006F007F008F009F010F011F014F015F016F017F018F020F021F022F023