DeepSeek R1-0528 Qwen3 8BOpen Weight

Released

DeepSeek R1-0528 Qwen3 8B is an AI model released by DeepSeek on Thursday, May 29 2025, 1 day after DeepSeek R1-0528. It is an open-weight model — the trained weights are available to download and run. It is a 8B parameterThe learned values that shape how a model responds. B means billion and T means trillion. For mixture-of-experts models, the total count is larger than the number active for each token. Parameter count alone does not measure quality. model. Benchmark results (shown below) cover LiveCodeBench, AIME 2024, AIME 2025, and GPQA Diamond.

Benchmarks

Coding

LiveCodeBenchCompetitive coding — Coding problems published so recently the AI can't have seen them in training — a contamination-free test of raw programming skill. Higher is better.
60.5%
#33 of 60

Reasoning & science

AIME 2024Reasoning — single-answer accuracy on the 2024 invitational mathematics exam without tools
86%
#8 of 21
AIME 2025Reasoning — single-answer accuracy on the 2025 invitational mathematics exam without tools
76.3%
#24 of 29
GPQA DiamondScience — Graduate-level science questions in biology, physics, and chemistry — hard enough that subject-matter PhDs score around 65%. Higher is better.
61.1%
#104 of 125

Compare DeepSeek R1-0528 Qwen3 8B with

DeepSeek R1-0528 Qwen3 8B

About

DeepSeek released this 8B distillation on May 29, 2025, following the R1-0528 update. It used reasoning traces from the larger model to post-train Qwen3 8B Base.

The smaller checkpoint had its own evaluation table and weights. Its scores here are taken from that table, keeping it separate from both full R1-0528 and the Qwen2.5-based distillations released in January.

Frequently asked questions

DeepSeek R1-0528 Qwen3 8B was released by DeepSeek on Thursday, May 29 2025.

All DeepSeek releases

30 tracked