Skip to main contentThe Local Bench

Math and reasoning · DeepSeek

DeepSeek R1-0528 Qwen3 8B

Hugging Face: deepseek-ai/DeepSeek-R1-0528-Qwen3-8B · checked 2026-10-13

R1-0528 reasoning traces distilled into Qwen3 8B. Long thinking, fits a 12 GB card at Q4.

Datasheet

Maker
DeepSeek
Parameters
8.19 BVerifiedsafetensors.total from the HF API Source
Licence
MITVerified
Access
Public download
Task (HF pipeline)
text-generation
First published on HF
2025-05-29
Last updated on HF
2025-05-29
Weights in memory
16 GB at 16-bit · 5.0 GB at ~4.85 bitsEstimateParameters × bits ÷ 8. Excludes context cache and runtime overhead; not every model has a 4-bit build.

Benchmark

86%
AIME 2024, pass@1 · #2 of 7 scored in this category

AIME 24 column, model card table. Source

Other published results

  • AIME 2025, pass@1: 76.3%Sourced Source

See the full math and reasoning ranking

The description summarises the model card at https://huggingface.co/deepseek-ai/DeepSeek-R1-0528-Qwen3-8B. Parameters and licence are read from the Hugging Face API; the weights-in-memory line is our arithmetic, not a measurement.