Qwen3 235B-A22B Instruct 2507 vs Qwen3.5 27B
| Spec | Qwen3 235B-A22B Instruct 2507 | Qwen3.5 27B |
|---|---|---|
| Respan Index | 20.2 (#117) | 20.2 (#117) |
| Blended price per 1M | $0.205 | $0.236 |
| Context window | 262K | 262K |
| Input / output per 1M | $0.09 / $0.55 | $0.086 / $0.688 |
| Max output | - | - |
| Input | text | text, vision |
| Released | 2025-07-21 | - |
Respan Index by category
Each category is scored 0 to 100, and how far ahead the winner is. Leads under a point are close to even.
| Category | Qwen3 235B-A22B Instruct 2507 | Qwen3.5 27B |
|---|---|---|
| Coding | 5.0 Qwen3 235B-A22B Instruct 2507 | 5.8 Qwen3.5 27B |
| Agents | 18.4* Qwen3 235B-A22B Instruct 2507 | 11.2 Qwen3.5 27B |
| Reasoning | 22.4* Qwen3 235B-A22B Instruct 2507 | 25.1 Qwen3.5 27B |
| Knowledge & language | 15.2 Qwen3 235B-A22B Instruct 2507 | 22.6 Qwen3.5 27B |
| Human preference | 48.1 Qwen3 235B-A22B Instruct 2507 | 49.3 Qwen3.5 27B |
* Estimated: no published scores in that category. How the Respan Index works
Benchmarks both report
Qwen3.5 27B leads on 2 of the 3 benchmarks both report. Higher is better. Each score links to where it was published.
| Benchmark | Qwen3 235B-A22B Instruct 2507 | Qwen3.5 27B |
|---|---|---|
| LMArena Elo | 1422 | 1409 |
| WeirdML (Håvard Tveit Ihle) | 38.7% | 39.5% |
| MMLU-Pro | 83% | 86.1% |
FAQ
Is Qwen3 235B-A22B Instruct 2507 better than Qwen3.5 27B?
Qwen3 235B-A22B Instruct 2507 and Qwen3.5 27B tie on the Respan Index at 20.2. Qwen3.5 27B leads on 2 of the 3 benchmarks both report.
Which is cheaper, Qwen3 235B-A22B Instruct 2507 or Qwen3.5 27B?
Qwen3 235B-A22B Instruct 2507 lists at $0.09 input and $0.55 output per 1M tokens; Qwen3.5 27B at $0.086 and $0.688. Qwen3 235B-A22B Instruct 2507 is 1.2x cheaper per token at list price.
Which has the larger context window?
Both accept 262K tokens of context.