AI Research

    Deepseek R1 0528 Qwen3 8b: gPQA Diamond is documented at 61.1% in DeepSeek’s published evaluation

    New benchmarks confirm DeepSeek's 8B parameter model achieves 61.1% on the rigorous GPQA Diamond test.

    GPQA Diamond is documented at 61.1% in DeepSeek’s published evaluation — verified by the OnAirToday AI Progress Tracker against the vendor's published documentation.

    O

    OnAirToday

    Newsroom

    2 min
    OnAirTodayView the primary source
    Deepseek R1 0528 Qwen3 8b: gPQA Diamond is documented at 61.1% in DeepSeek’s published evaluation

    The OnAirToday AI Progress Tracker recorded a verified change to Deepseek R1 0528 Qwen3 8b (version deepseek-r1-0528-qwen3-8b): GPQA Diamond is documented at 61.1% in DeepSeek’s published evaluation.

    DeepSeek-R1-0528-Qwen3-8B is DeepSeek’s open-weight reasoning model built by fine-tuning Qwen3-8B on reasoning generated by DeepSeek-R1-0528. Its published evaluations show strengths in mathematics, science questions and coding, with substantially lower deployment requirements than the full-size teacher model.

    What changed in this review:

    • Added published AIME 2024, AIME 2025 and LiveCodeBench v6 results.
    • Documented a 131,072-token context configuration.
    • Added architecture, licensing, deployment guidance and comparisons with the earlier 7B distilled model.
    • Added the January 2025 DeepSeek service database exposure, explicitly distinguished from a vulnerability in this checkpoint.
    • Capability score 23 → 71
    • Flaws and limitations updated

    Teams building on Deepseek R1 0528 Qwen3 8b should confirm limits, pricing and benchmark coverage against the vendor's own documentation before relying on these figures.

    The full profile — capabilities, benchmark citations, known flaws and risk guidance — is on the tracker at /ai-models/deepseek-r1-0528-qwen3-8b.

    References:

    • https://huggingface.co/deepseek-ai/DeepSeek-R1-0528-Qwen3-8B
    • https://huggingface.co/deepseek-ai/DeepSeek-R1-0528-Qwen3-8B/blob/main/config.json
    • https://github.com/deepseek-ai/DeepSeek-R1
    • https://www.wiz.io/blog/wiz-research-uncovers-exposed-deepseek-database-leak
    AI Analysis

    Generated insight by OnAirToday AI

    The performance of the DeepSeek R1 0528 model on the GPQA Diamond benchmark demonstrates a significant leap in specialized reasoning capabilities for compact open-weight architectures. This milestone suggests that high-tier analytical proficiency is becoming increasingly accessible outside of proprietary, large-scale compute environments.

    Powered by AI • Not a substitute for human analysisMore Insights →
    View the primary source →

    Get the day's AI news in one short email. Free.

    Share
    Filed under:AI Research

    Subscribe to our Free AI Newsletter

    The Daily AI Digest, every morning. Pick your topics after you subscribe.

    Free. Unsubscribe anytime.

    Comments

    Keep comments relevant and respectful. No links, advertising, self-promotion, or spam.

    0/1000