standardMIT8B paramsOpen weights

DeepSeek-R1 8B (0528 Qwen3)

8B parameter open-weight model. Qwen3-8B distillation of the R1-0528 update — Ollama's deepseek-r1:8b tag now points at this checkpoint, replacing the original Llama-3 distill. Improved math, coding, and logic over the first-release distills. Thinking chains visible in output. Q4 fits in 8 GB VRAM.

DeepSeek · DeepSeek-R1

Model overview

DeepSeek-R1 8B (0528 Qwen3) is a distilled reasoning model that compresses DeepSeek's R1-0528 update into an 8B Qwen3-based checkpoint small enough to run on a single consumer GPU. It's a newer, improved distillation over the original Llama-3-based R1 8B distill, with better math, coding, and logical reasoning, and it exposes its intermediate "thinking" steps in its output so you can inspect how it arrived at an answer. This makes it a practical entry point for experimenting with chain-of-thought reasoning behavior without needing server-class hardware, though as with any distilled model it trades some depth versus the full-size R1 for that efficiency.

Check your GPU fitWhat is VRAM?What is quantization?Local LLM quickstart

Editorial review

Reviewed byOpenSourcesAI EditorialLast updatedJune 2026SourcesHuggingFace model card (deepseek-ai/DeepSeek-R1-0528-Qwen3-8B), official docs, OpenSourcesAI editorial review.

VRAM figures are empirical estimates. Actual usage varies by runtime, context length, and system configuration. Verify on your specific hardware before production use.

Will DeepSeek-R1 8B (0528 Qwen3) run on your machine?

DeepSeek-R1 8B (0528 Qwen3) is 8B parameters and needs 6.5 GB of VRAM at Q4_K_M — 5 GB of weights plus 1.5 GB of runtime overhead for the inference server itself.

VRAM by quantization

QuantizationWeightsNeeds (with overhead)Quality
Q4_K_M5 GB6.5 GBgood
Q8_08.8 GB10.3 GBhigh
FP1616 GB17.5 GBreference

Fit on common hardware at Q4_K_M

HardwareMemory the model can useSystem RAMVerdict
CPU OnlyNone (CPU only)16 GBCPU offload
RTX 4060 Laptop8 GB16 GBTight
RTX 3060 (12GB)12 GB32 GBComfortable
RTX 4060 Ti (16GB)16 GB32 GBComfortable
RTX 309024 GB64 GBComfortable

Comfortable means VRAM clears the requirement by 2 GB or more. Tight means it covers the requirement with no margin. CPU offload means the model does not fit in VRAM but system RAM is at least 1.6× the weights, so it will run at reduced speed — expect roughly 1–5 tokens per second. Figures are weights plus a fixed runtime overhead and exclude KV-cache growth, which scales with context length.

Check DeepSeek-R1 8B (0528 Qwen3) against your exact hardwareHow these verdicts are calculated

Need more hardware for DeepSeek-R1 8B (0528 Qwen3)? Open the PC Builder for the 7B / 8B tier →

Related models

Llama 3 8B8B parameter model — shared focus: chat, coding, rag.Qwythos-9B (Claude-Mythos-5 1M)9B parameter model — shared focus: reasoning, coding, tool-use.Qwen3.5 9B9.7B parameter model — shared focus: chat, reasoning, long-context.DeepSeek Coder 33BCoding-model comparisons and legacy baseline testing.

Ready to run this model locally?

Find a compatible interface in our Local AI Tools directory →