Across 14 shared benchmarks, Qwen 3.6 Plus Preview leads overall: Qwen 3.6 Plus Preview wins 12, Qwen3.5-397B-A17B wins 2, with 0 ties and an average score difference of +2.59.
阿里巴巴 · 2026-03-31 · AI model
阿里巴巴 · 2026-02-16 · Multimodal model
Grouped by capability, sorted by largest gap within each. 14 shared benchmarks.
| Benchmark | Qwen 3.6 Plus Preview | Qwen3.5-397B-A17B | Diff |
|---|---|---|---|
| SWE-Bench Pro - Public | 56.609 / 36Thinking (With Tools) | 50.9023 / 36Thinking (No Tools) | +5.70 |
| SWE-bench Multilingual | 73.805 / 17Thinking (No Tools) | 69.3014 / 17Thinking (No Tools) | +4.50 |
| LiveCodeBench | 87.109 / 118Thinking (No Tools) | 83.6018 / 118Thinking (No Tools) | +3.50 |
| SWE-bench Verified | 78.8016 / 103Thinking (With Tools) | 76.4024 / 103Thinking (With Tools) | +2.40 |
| Benchmark | Qwen 3.6 Plus Preview | Qwen3.5-397B-A17B | Diff |
|---|---|---|---|
| HLE | 50.6014 / 149Thinking (With Tools) | 48.3023 / 149Thinking (With Tools + Internet) | +2.30 |
| GPQA Diamond | 90.4015 / 175Thinking (No Tools) | 88.4023 / 175Thinking (No Tools) | +2 |
| MMLU Pro | 88.504 / 124Thinking (No Tools) | 87.808 / 124Thinking (No Tools) | +0.70 |
| C-Eval |
| Benchmark | Qwen 3.6 Plus Preview | Qwen3.5-397B-A17B | Diff |
|---|---|---|---|
| Terminal Bench 2.0 | 61.6014 / 43Thinking (With Tools) | 52.5026 / 43Thinking (With Tools) | +9.10 |
| Tool Decathlon | 39.804 / 7Thinking (With Tools) | 38.305 / 7Thinking (With Tools) | +1.50 |
| Benchmark | Qwen 3.6 Plus Preview | Qwen3.5-397B-A17B | Diff |
|---|---|---|---|
| AIME 2026 | 95.302 / 14Thinking (No Tools) | 91.3011 / 14Thinking (No Tools) | +4 |
| IMO-AnswerBench | 83.809 / 17Thinking (No Tools) | 80.9013 / 17Thinking (No Tools) | +2.90 |
| Benchmark | Qwen 3.6 Plus Preview | Qwen3.5-397B-A17B | Diff |
|---|---|---|---|
| IF Bench | 74.205 / 27Thinking (No Tools) | 76.502 / 27Thinking (No Tools) | -2.30 |
| Benchmark | Qwen 3.6 Plus Preview | Qwen3.5-397B-A17B | Diff |
|---|---|---|---|
| AA-LCR | 68.306 / 13Thinking (No Tools) | 68.705 / 13Thinking (No Tools) | -0.40 |
| Field | Qwen 3.6 Plus Preview | Qwen3.5-397B-A17B |
|---|---|---|
| Publisher | 阿里巴巴 | 阿里巴巴 |
| Release date | 2026-03-31 | 2026-02-16 |
| Model type | AI model | Multimodal model |
| Architecture | Dense | MoE |
| Parameters | 0.0 | 397.0 |
| Context length | 1M | 256K |
| Max output | 65536 | Not available |
Prices use DataLearner records when available; missing fields are not inferred.
| Item | Qwen 3.6 Plus Preview | Qwen3.5-397B-A17B |
|---|---|---|
| Text input | $0.5 / 1M tokens | $0.5 / 1M tokens |
| Text output | $3 / 1M tokens | $3 / 1M tokens |
| Cache read | $0.05 / 1M tokens | $0.05 / 1M tokens |
| Cache write | $0.625 / 1M tokens | $0.625 / 1M tokens |
On average across the 14 shared benchmarks, Qwen 3.6 Plus Preview scores 2.59 higher.
Largest single-benchmark gap: Terminal Bench 2.0 — Qwen 3.6 Plus Preview 61.60 vs Qwen3.5-397B-A17B 52.50 (+9.10).
Page generated from structured model, pricing and benchmark records. No real-time LLM is used to write the prose.
| 93.302 / 9Thinking (No Tools) |
| 933 / 9Thinking (No Tools) |
| +0.30 |
