RLTasksby ulam.ai
Models / Qwen

Qwen / MODEL EVIDENCE

Qwen3.8 Max (unknown)

5 benchmarks · 6 observations · 5 sources · released 2026-08-02

Scores at a glance

5 benchmarks · large dot: Qwen3.8 Max (unknown); small dots: closest models where reported

CritPt ↗20%
032.29

Qwen3.8 Max (unknown) · CritPt · Accuracy · % ↗

Closest models by score

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

6.5 percentile-point gap

3 of 5 shared benchmark comparisons

6.9 percentile-point gap

3 of 5 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 Qwen3.8 Max / 2.4T-A95B report says

The official model card identifies 2.4T-A95B as post-trained and distinguishes its open weights from the Max service, which adds features. It documents the release and architecture without a reproducible optimization recipe.

  • Post-trained open-weight checkpoint
  • Separate Max service configuration
Supervised fine-tuning · background →Preference learning · background →Stateful agent training · background →

No named optimizer, reward mixture or stage-by-stage training schedule is specified in this card. The method links are background, not claims of adoption; open weights and the Max service remain separate model identities.

  1. Qwen3.8-2.4T-A95B 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. SciCode Benchmark | Artificial Analysis ↗

    Artificial Analysis

  3. APEX-Agents ↗

    Epoch AI

  4. CritPt v1 ↗

    Epoch AI

  5. ProofBench — evaluation methodology and results ↗

    Vals AI · 2026

Names used in source tables

Qwen3.8 Max (unknown)