deepseek-ai / TEXT MODEL

DeepSeek-R1-Distill-Qwen-14B

A Qwen-based distilled reasoning model trained using DeepSeek-R1 outputs for mathematical, coding and reasoning tasks.

Source reviewed 2026-10-24 · AiBrainWorX editorial summary

Creator: deepseek-ai
Parameters: 14.77B (HF tensor count)
Context: 131,072 tokens
License: mit

01 / CAPABILITIES

What it is built for

Reasoning-focused text generation
Qwen-based architecture
Distilled from DeepSeek-R1 outputs

Inputs: Text

Configuration value; backend support and enough memory are required. Start with a smaller context on limited hardware.

02 / ON YOUR MACHINE

Mid-range GPU / CPU offload

Try logic puzzles. Reasoning can consume the output budget before a final answer; tiny caps may yield no usable answer.

Choose an explicitly supported quantization in Ollama, LM Studio or another compatible local server.
Start the backend with a context that fits your machine. Strata supports its own specific architecture, not every model in this library.
In the launcher, select that backend, refresh its inventory and save the exact loaded model identifier.

For stories and games, try tool-free Local chat. Coding agents additionally require reliable tool calling and the correct backend parser.

Launcher integration

Connect a supported local text/vision backend; not independently tested

Runtime and training workflow

Supported local text backend

Separate model-specific training workflow. Training suitability requires exact checkpoint format and prepared data; a catalog entry is not a completed training test.

Compare dataset preparation guides →

03 / WHAT IS VERIFIED

Evidence, with context

Our test status

Source reviewed; no local generation benchmark for this exact checkpoint.

We have not added a benchmark number without documenting its test conditions. A model-card capability is not a measured speed or quality guarantee on your hardware.

04 / OWNERSHIP & USE

mit

Keep upstream notices and check the exact downloaded artifact and parent-model terms. The original creator remains credited.

AiBrainWorX authors this profile and the launcher. deepseek-ai remains the model’s creator. Models are not rebranded as our own weights.

05 / GO DEEPER

Original sources & model files

This page is our concise summary, not a mirror of the full upstream card. Review the original documentation and exact artifact before downloading.

Official card on Hugging Face ↗Files & versions ↗Reviewed card revision ↗

Reviewed revision: 1df8507178af · Specifications may change after the review date.