AI Progress Tracker

    Qwen3 4b Instruct · Qwen3-4B-Instruct-2507

    6 Aug 2025

    Ask the Advisor about this model
    48from 2
    benchmarks

    Qwen3-4B-Instruct-2507 is an open-weight, text-only instruction model from Alibaba's Qwen team, with native support for 262,144-token contexts. It uses a non-thinking configuration and targets multilingual assistance, coding, reasoning and tool-use applications.

    context: 262Ktext

    Benchmark results

    Every figure links to where it was published. Compare like with like — harnesses and prompt setups differ between vendors.

    BenchmarkScoreSource
    GPQA DiamondPhD-level science questions45.8%28 Aug 2026View
    OTIS Mock AIME 2024-25Competition mathematics52.2%28 Aug 2026View

    Weighted mean of 2 verified benchmark results: GPQA Diamond 45.8%, OTIS Mock AIME 2024-25 52.2%.

    Capabilities

    • Instruction-following and general-purpose text generation.
    • Multilingual assistance and long-tail knowledge coverage.
    • Mathematical reasoning and coding assistance.
    • Tool-use workflows through compatible application frameworks.
    • Long-document processing within a native 262,144-token context window.

    What improved

    • The vendor reports stronger instruction following, reasoning, mathematics, coding and tool use than the earlier Qwen3-4B non-thinking configuration.
    • The vendor reports broader long-tail knowledge coverage across multiple languages.
    • The vendor reports better alignment with user preferences in subjective and open-ended tasks.
    • Native context support increases to 262,144 tokens.

    Known flaws & limitations

    • Supports text input and output, not native image or audio understanding.
    • Does not support a switchable thinking mode or produce thinking blocks.
    • Requires software with Qwen3 architecture support; Transformers versions earlier than 4.51.0 can fail to load it.

    Risks, safety & security issues

    • Not yet confirmed by public reporting.

    Good fit for

    • Self-hosted, text-based assistants using a compact open-weight model.
    • Multilingual drafting, rewriting and instruction-following tasks.
    • Long-document summarization and question answering.
    • Coding assistance and application-managed tool-use workflows.

    Poor fit for

    • Applications requiring native image, audio or video understanding.
    • Workflows requiring a selectable thinking mode.
    • Deployments using inference software without Qwen3 architecture support.
    Security & risk dashboardLatest vulnerabilities, advisories and real-world incidents for Qwen3 4b Instruct

    References

    Profile last reviewed 10/18/2026 and re-checked automatically every few hours against the vendor's published documentation, papers and public leaderboards. Scores come from the cited benchmark results above — confirm limits and pricing against the provider's own documentation before building on them.