Qwen·released 2025-07-251 source
Qwen3-235B-A22B-Instruct (Jul 2025) benchmark scores: 7 benchmarks tracked. 0% of its scores are independently reproduced — the rest are self-reported or unverified.
Leads on 0 of 7 benchmarks·best result 64 on DTBench (unverified)·0 of 7 independently reproduced
| Benchmark | Score | Evidence status | Measured |
|---|---|---|---|
| DTBench | 64accuracy (%) | unverifiedT1 | 2025-07-25https://conceptualreasoning.ai/dtbench |
| Aider polyglot | 59.6accuracy (%) | unverified· optimizedT1 | 2025-07-25https://aider.chat/docs/leaderboards/ |
| Fiction.LiveBench | 52.9accuracy (%) | unverifiedT1 | 2025-07-25Fiction.live leaderboard |
| WeirdML | 38.7accuracy (%) | unverifiedT1 | 2025-07-25WeirdML Leaderboard |
| LMCA | 27.67accuracy (%) | unverifiedT1 | 2025-07-25https://conceptualreasoning.ai/lmca |
| ARC-AGI | 11accuracy (%) | unverified· optimizedT2 | 2025-07-25 |
| ARC-AGI-2 | 1.25accuracy (%) | unverifiedT2 | 2025-07-25 |
Claims drawn from cited facts, not live model generation.
Developed by Qwen and released in 2025, this model sits at the large end of the spectrum with 235 billion parameters, a scale that reflects a considerable investment in capacity rather than a lean, budget-conscious design. That size buys substantially more headroom for complex tasks, but it comes paired with heavier compute demands and higher serving costs than a compact model would carry.
4 cited facts
This model is scored on 7 tracked benchmarks, and it holds no current top score on any of them. Its record is unverified: neither confirmed nor disputed by any outside evaluation, so the reported figures should be read as claims awaiting independent check.
3 cited facts
This model trails the state of the art by an average of 55.47 points across its benchmarks, a wide gap that under the disclosed harness places it well behind the leaders. Consistently, it holds the top score on 0 benchmarks, so there is no category in which it currently leads.
2 cited facts
2 facts cross-checked across data sources: 2 single-source.
Source facts, citations, and refresh stamp for this record.
Sources: curated_param_counts, epoch_benchmarks
Qwen3-235B-A22B-Instruct (Jul 2025) is an AI model developed by Qwen, released 2025-07-25. Its benchmark record below tags every score as independently reproduced, self-reported, or unverified.
Qwen3-235B-A22B-Instruct (Jul 2025) has recorded scores on 7 benchmarks, each shown with its evidence status.
Qwen3-235B-A22B-Instruct (Jul 2025) has recorded scores on 7 benchmarks — DTBench, WeirdML, LMCA, Aider polyglot, ARC-AGI, Fiction.LiveBench, and 1 more. The full table above shows each score with its evidence status.
0 of 7 recorded scores (0%) are independently reproduced rather than self-reported by the lab.
Qwen3-235B-A22B-Instruct (Jul 2025) has 7 tracked claims: 7 unverified.