Model comparison
Qwen3.5 397B vs Qwen3.6 Plus
Head-to-head evidence from 53 shared benchmark results across 8 categories. Overall scores shown here use the public BenchAlign v5 ranking lane.
Public leaderboard positions: Qwen3.5 397B #71 (Estimated); Qwen3.6 Plus #30 (Supported). Intervals and evidence labels describe ranking uncertainty, not a guarantee for a specific workload.
Evidence parity. Qwen3.5 397B and Qwen3.6 Plus share 53 comparable benchmark results. 8 of 8 categories are comparable. 2 results are unique to Qwen3.5 397B; 7 to Qwen3.6 Plus.
Updated July 23, 2026- Shared results
- 53
- Qwen3.5 397B only
- 2
- Qwen3.6 Plus only
- 7
- Comparable categories
- 8 / 8
Pick Qwen3.6 Plus if you want the stronger benchmark profile. Qwen3.5 397B only becomes the better choice if mathematics is the priority or you would rather avoid the extra latency and token burn of a reasoning model.
Why this result
Qwen3.6 Plus is clearly ahead on the BenchAlign aggregate, 65.2 to 57.01. The gap is large enough that you do not need to squint at the spreadsheet to see the difference.
Qwen3.6 Plus's sharpest advantage is in agentic, where it averages 61.6 against 56.5. The single biggest benchmark swing on the page is Terminal-Bench 2.0, 52.5% to 61.6%. Qwen3.5 397B does hit back in mathematics, so the answer changes if that is the part of the workload you care about most.
Qwen3.6 Plus is the reasoning model in the pair, while Qwen3.5 397B is not. That usually helps on harder chain-of-thought-heavy tests, but it can also mean more latency and more token spend in real use. Qwen3.6 Plus gives you the larger context window at 1M, compared with 128K for Qwen3.5 397B.
Category breakdown
Exact category averages are shown below. Not measured means BenchLM does not have enough sourced public coverage for that model and category.
| Category | Qwen3.5 397B | Δ | Qwen3.6 Plus |
|---|---|---|---|
| Math | Qwen3.5 397B90.6 | Margin← 30.1 | Qwen3.6 Plus60.5 |
| Inst. Following | Qwen3.5 397B92.6 | Margin← 10.3 | Qwen3.6 Plus82.3 |
| Agentic | Qwen3.5 397B56.5 | Margin→ 5.1 | Qwen3.6 Plus61.6 |
| Coding | Qwen3.5 397B66.5 | Margin→ 3.8 | Qwen3.6 Plus70.3 |
| Reasoning | Qwen3.5 397B63.2 | Margin← 1.2 | Qwen3.6 Plus62.0 |
| Knowledge | Qwen3.5 397B56.6 | Margin→ 0.5 | Qwen3.6 Plus57.1 |
| Multimodal | Qwen3.5 397B79.6 | Margin→ 0.2 | Qwen3.6 Plus79.8 |
| Multilingual | Qwen3.5 397B84.7 | MarginTie | Qwen3.6 Plus84.7 |
Decisive benchmark drivers
The largest measured benchmark gaps in this matchup, with exact reported values.
More
- Source ↗
Terminal-Bench 2.0
AgenticA 52.5%B 61.6%Winner: Qwen3.6 PlusΔ 9.1Terminal-Bench 2.0: Qwen3.5 397B scored 52.5%; Qwen3.6 Plus scored 61.6%. Qwen3.6 Plus wins this benchmark. - Source ↗
SWE-bench Pro
CodingA 50.9%B 56.6%Winner: Qwen3.6 PlusΔ 5.7SWE-bench Pro: Qwen3.5 397B scored 50.9%; Qwen3.6 Plus scored 56.6%. Qwen3.6 Plus wins this benchmark. - Source ↗
SWE-bench Verified
CodingA 76.2%B 78.8%Winner: Qwen3.6 PlusΔ 2.6SWE-bench Verified: Qwen3.5 397B scored 76.2%; Qwen3.6 Plus scored 78.8%. Qwen3.6 Plus wins this benchmark. - Source ↗
GPQA
KnowledgeA 88.4%B 90.4%Winner: Qwen3.6 PlusΔ 2GPQA: Qwen3.5 397B scored 88.4%; Qwen3.6 Plus scored 90.4%. Qwen3.6 Plus wins this benchmark. - Source ↗
AIME26
MathA 93.3%B 95.3%Winner: Qwen3.6 PlusΔ 2AIME26: Qwen3.5 397B scored 93.3%; Qwen3.6 Plus scored 95.3%. Qwen3.6 Plus wins this benchmark.
Operational comparison
Runtime and commercial metrics are compared only when both models have a complete sourced value.
| Metric | Qwen3.5 397B | Qwen3.6 Plus | Comparison |
|---|---|---|---|
| Input / output priceUSD per 1M tokens | Qwen3.5 397B$0.6 input / $3.6 output | Qwen3.6 PlusNot available | A complete price comparison is not available. |
| Generation speedtokens per second | Qwen3.5 397B96 tok/s | Qwen3.6 PlusNot available | A complete speed comparison is not available. |
| First-answer latencyseconds to first token | Qwen3.5 397B2.44 s | Qwen3.6 PlusNot available | A complete latency comparison is not available. |
| Context windowmaximum listed tokens | Qwen3.5 397B128K | Qwen3.6 Plus1M | Qwen3.6 Plus lists the larger context window. |
Benchmark Deep Dive
AgenticQwen3.6 Plus wins18 benchmarks
| Benchmark | Qwen3.5 397B | Qwen3.6 Plus | Result |
|---|---|---|---|
| Terminal-Bench 2.0Source | 52.5% | 61.6% | Qwen3.6 Plus leads |
| BrowseCompSource | 62% | — | Not comparable |
| Claw-EvalSource | 56.8% | 58.8% | Qwen3.6 Plus leads |
| QwenClawBenchSource | 51.8% | 57.2% | Qwen3.6 Plus leads |
| τ³-bench resultsSource | 68.4% | 70.7% | Qwen3.6 Plus leads |
| VITA-BenchSource | 43.7% | 44.3% | Qwen3.6 Plus leads |
| DeepPlanningSource | 37.6% | 41.5% | Qwen3.6 Plus leads |
| ToolathlonSource | 36.3% | 39.8% | Qwen3.6 Plus leads |
| MCP AtlasSource | 46.1% | 48.2% | Qwen3.6 Plus leads |
| MCP-TasksSource | 74.2% | 74.1% | Qwen3.5 397B leads |
| WideResearchSource | 74.0% | 74.3% | Qwen3.6 Plus leads |
| τ²-bench resultsSource | 95.6% | 97.7% | Qwen3.6 Plus leads |
| Gert LabsSource | 46.76% | 50.60% | Qwen3.6 Plus leads |
| ResearchClawBenchSource | 14.2% | 18.0% | Qwen3.6 Plus leads |
| AA Agentic IndexSource | 19.9% | 27.6% | Qwen3.6 Plus leads |
| APEX-Agents-AASource | 15.3% | — | Not comparable |
| GDPval-AASource | 23.1% | 31.8% | Qwen3.6 Plus leads |
| GDPval-AASource | 962 | 1135 | Qwen3.6 Plus leads |
CodingQwen3.6 Plus wins7 benchmarks
| Benchmark | Qwen3.5 397B | Qwen3.6 Plus | Result |
|---|---|---|---|
| SWE-bench VerifiedSource | 76.2% | 78.8% | Qwen3.6 Plus leads |
| LiveCodeBench v6Source | 83.6% | 87.1% | Qwen3.6 Plus leads |
| SWE-bench ProSource | 50.9% | 56.6% | Qwen3.6 Plus leads |
| AA-SciCodeSource | 42.0% | 40.7% | Qwen3.5 397B leads |
| AA Coding IndexSource | 48.2% | 54.5% | Qwen3.6 Plus leads |
| SWE MultilingualSource | — | 73.8% | Not comparable |
| Vibe Code BenchSource | — | 25.56% | Not comparable |
ReasoningQwen3.5 397B wins4 benchmarks
KnowledgeQwen3.6 Plus wins12 benchmarks
| Benchmark | Qwen3.5 397B | Qwen3.6 Plus | Result |
|---|---|---|---|
| GPQASource | 88.4% | 90.4% | Qwen3.6 Plus leads |
| SuperGPQASource | 70.4% | 71.6% | Qwen3.6 Plus leads |
| MMLU-ProSource | 87.8% | 88.5% | Qwen3.6 Plus leads |
| MMLU-ReduxSource | 94.9% | 94.5% | Qwen3.5 397B leads |
| C-EvalSource | 93% | 93.3% | Qwen3.6 Plus leads |
| HLESource | 28.7% | 28.8% | Qwen3.6 Plus leads |
| Artificial Analysis Intelligence IndexSource | 33.7% | 39.6% | Qwen3.6 Plus leads |
| AA-GPQA DiamondSource | 89.3% | 88.2% | Qwen3.5 397B leads |
| AA-HLESource | 27.3% | 25.7% | Qwen3.5 397B leads |
| AA-Omniscience IndexSource | -29.8% | 2.7% | Qwen3.6 Plus leads |
| AA-Omniscience AccuracySource | 31.4% | 26.2% | Qwen3.5 397B leads |
| AA-Omniscience Hallucination RateSource | 89.1% | 32.0% | Qwen3.6 Plus leads |
MathQwen3.5 397B wins7 benchmarks
| Benchmark | Qwen3.5 397B | Qwen3.6 Plus | Result |
|---|---|---|---|
| AIME26Source | 93.3% | 95.3% | Qwen3.6 Plus leads |
| HMMT Feb 2025Source | 94.8% | 96.7% | Qwen3.6 Plus leads |
| HMMT Nov 2025Source | 92.7% | 94.6% | Qwen3.6 Plus leads |
| HMMT Feb 2026Source | 87.9% | 87.8% | Qwen3.5 397B leads |
| MMAnswerBenchSource | 80.9% | 83.8% | Qwen3.6 Plus leads |
| FrontierMath v2 (Tiers 1-3)Source | — | 26.207% | Not comparable |
| FrontierMath v2 (Tier 4)Source | — | 8.333% | Not comparable |
MultilingualTie2 benchmarks
MultimodalQwen3.6 Plus wins9 benchmarks
| Benchmark | Qwen3.5 397B | Qwen3.6 Plus | Result |
|---|---|---|---|
| MMMU-ProSource | 79% | 78.8% | Qwen3.5 397B leads |
| MathVisionSource | 88.6% | 88.0% | Qwen3.5 397B leads |
| CharXivSource | 80.8% | 81.5% | Qwen3.6 Plus leads |
| VideoMMMUSource | 84.7% | 84.0% | Qwen3.5 397B leads |
| ScreenSpot ProSource | 65.6% | 68.2% | Qwen3.6 Plus leads |
| V*Source | 95.8% | 96.9% | Qwen3.6 Plus leads |
| AA-MMMU-ProSource | 77.3% | 78.0% | Qwen3.6 Plus leads |
| MMMUSource | — | 86.0% | Not comparable |
| Design Arena WebsiteSource | — | 1249 | Not comparable |
Frequently Asked Questions (9)
Which is better, Qwen3.5 397B or Qwen3.6 Plus?
Qwen3.6 Plus is ahead on BenchLM's BenchAlign leaderboard, 65.2 to 57.01. The biggest single separator in this matchup is Terminal-Bench 2.0, where the scores are 52.5% and 61.6%.
Which is better for knowledge tasks, Qwen3.5 397B or Qwen3.6 Plus?
Qwen3.6 Plus has the edge for knowledge tasks in this comparison, averaging 57.1 versus 56.6. Inside this category, AA-Omniscience Hallucination Rate is the benchmark that creates the most daylight between them.
Which is better for coding, Qwen3.5 397B or Qwen3.6 Plus?
Qwen3.6 Plus has the edge for coding in this comparison, averaging 70.3 versus 66.5. Inside this category, AA Coding Index is the benchmark that creates the most daylight between them.
Which is better for math, Qwen3.5 397B or Qwen3.6 Plus?
Qwen3.5 397B has the edge for math in this comparison, averaging 90.6 versus 60.5. Inside this category, MMAnswerBench is the benchmark that creates the most daylight between them.
Which is better for reasoning, Qwen3.5 397B or Qwen3.6 Plus?
Qwen3.5 397B has the edge for reasoning in this comparison, averaging 63.2 versus 62. Inside this category, AA-LCR is the benchmark that creates the most daylight between them.
Which is better for agentic tasks, Qwen3.5 397B or Qwen3.6 Plus?
Qwen3.6 Plus has the edge for agentic tasks in this comparison, averaging 61.6 versus 56.5. Inside this category, GDPval-AA is the benchmark that creates the most daylight between them.
Which is better for multimodal and grounded tasks, Qwen3.5 397B or Qwen3.6 Plus?
Qwen3.6 Plus has the edge for multimodal and grounded tasks in this comparison, averaging 79.8 versus 79.6. Inside this category, ScreenSpot Pro is the benchmark that creates the most daylight between them.
Which is better for instruction following, Qwen3.5 397B or Qwen3.6 Plus?
Qwen3.5 397B has the edge for instruction following in this comparison, averaging 92.6 versus 82.3. Inside this category, AA-IFBench is the benchmark that creates the most daylight between them.
Which is better for multilingual tasks, Qwen3.5 397B or Qwen3.6 Plus?
Qwen3.5 397B and Qwen3.6 Plus are effectively tied for multilingual tasks here, both landing at 84.7 on average.
Related Comparisons
Explore More
Choose a model with this week’s evidence
Join 2,000+ readers for ranking moves, pricing changes, and the claims that still need proof.
One email each week. Unsubscribe anytime.