RLTasksby ulam.ai
Models / Qwen

Qwen / MODEL EVIDENCE

Qwen3.5 Plus

25 benchmarks · 27 observations · 26 sources · released 2026-02-16

Scores at a glance

25 benchmarks · large dot: Qwen3.5 Plus; small dots: closest models where reported

LMCA ↗36.36%
2.7868.26

Qwen3.5 Plus · LMCA · Score · % ↗

DTBench ↗80.53%
41.6598.93

Qwen3.5 Plus · DTBench · Accuracy · % ↗

Closest models by score

Similar positions on shared benchmark leaderboards. Smaller gaps are closer.

14.2 percentile-point gap

14 of 18 shared benchmark comparisons

15.7 percentile-point gap

11 of 18 shared benchmark comparisons

How similarity is measured

Mean absolute gap in percentile positions on shared benchmark settings; lower is closer. At least 60% overlap is required, with a minimum of three when available. Independent evaluations are used when there are at least three; otherwise published report comparisons are used. Repeated model observations within one setting are averaged before ranking. Writing-style traits and composite indices are excluded. This describes similarity, not an overall quality ranking.

POST-TRAINING / SOURCE DISCLOSURE

What the Qwen3.5 post-trained family report says

The reviewed card describes progressively harder agent environments and asynchronous reinforcement-learning infrastructure. Its architecture and multimodal training are documented separately from its post-training objective.

  • Agent environments with increasing task complexity
  • Large-scale asynchronous reinforcement learning
Stateful agent training →RL training systems · background →

The card reviewed is for 397B-A17B. It does not name a complete policy objective; the background links do not establish that every Qwen3.5 checkpoint used the same optimizer.

  1. Qwen3.5-397B-A17B model card ↗

    Qwen Team · 2026

Filter comparisons to this model →

Sources

Every plotted observation links to its evaluation. Reported configurations, benchmark versions and task subsets can differ; similarity describes shared scores, not training lift.

  1. Arena leaderboard — WebDev evaluation ↗

    Arena

  2. DTBench — Conceptual Reasoning Index ↗

    Conceptual Reasoning Index

  3. LMCA — Conceptual Reasoning Index ↗

    Conceptual Reasoning Index

  4. Chess Puzzles ↗

    Epoch AI

  5. CL-bench ↗

    Epoch AI

  6. CL-bench Life ↗

    Epoch AI

  7. FrontierMath Tier 4 (v1, old) ↗

    Epoch AI

  8. FrontierMath Tiers 1-3 (v1, old) ↗

    Epoch AI

  9. GPQA Diamond ↗

    Epoch AI

  10. Mystery Game Puzzles ↗

    Epoch AI

  11. OTIS Mock AIME 2024-2025 ↗

    Epoch AI

  12. SimpleQA Verified ↗

    Epoch AI

  13. Vending Bench 2 — methodology and published results ↗

    Epoch AI · 2026

  14. AIME — evaluation methodology and results ↗

    Vals AI · 2026

  15. CaseLaw — evaluation methodology and results ↗

    Vals AI · 2026

  16. CorpFin — evaluation methodology and results ↗

    Vals AI · 2026

  17. GPQA Diamond — evaluation methodology and results ↗

    Vals AI · 2026

  18. LiveCodeBench — evaluation methodology and results ↗

    Vals AI · 2026

  19. LegalBench — evaluation methodology and results ↗

    Vals AI · 2026

  20. MedQA — evaluation methodology and results ↗

    Vals AI · 2026

  21. MMLU Pro — evaluation methodology and results ↗

    Vals AI · 2026

  22. MMMU Pro — evaluation methodology and results ↗

    Vals AI · 2026

  23. MortgageTax — evaluation methodology and results ↗

    Vals AI · 2026

  24. SAGE — evaluation methodology and results ↗

    Vals AI · 2026

  25. SWE-bench Verified — evaluation methodology and results ↗

    Vals AI · 2026

  26. Vibe Code Bench — evaluation methodology and results ↗

    Vals AI · 2026

Names used in source tables

Qwen 3.5 Plus · Qwen 3.5 Plus (Mini-SWE-agent) · Qwen 3.5 Plus (OpenHands) · Qwen3.5 Plus