Model comparison
GPT-5.6 Luna vs Qwen3.5 397B
Head-to-head evidence from 21 shared benchmark results across 5 categories. Overall scores shown here use the public BenchAlign v5 ranking lane.
Public leaderboard positions: GPT-5.6 Luna #22 (Estimated); Qwen3.5 397B #71 (Estimated). Intervals and evidence labels describe ranking uncertainty, not a guarantee for a specific workload.
Evidence parity. GPT-5.6 Luna and Qwen3.5 397B share 21 comparable benchmark results. 5 of 8 categories are comparable. 20 results are unique to GPT-5.6 Luna; 34 to Qwen3.5 397B.
Updated July 23, 2026- Shared results
- 21
- GPT-5.6 Luna only
- 20
- Qwen3.5 397B only
- 34
- Comparable categories
- 5 / 8
Pick GPT-5.6 Luna if you want the stronger benchmark profile. Qwen3.5 397B only becomes the better choice if mathematics is the priority or you want the cheaper token bill.
Confidence note. This is a partial-evidence comparison with 21 shared benchmark results across 5 evidence categories; 5 of 8 categories currently have scoreable aggregates for both models. Treat the verdict as directional until coverage is more balanced.
Why this result
GPT-5.6 Luna is clearly ahead on the BenchAlign aggregate, 67.17 to 57.01. The gap is large enough that you do not need to squint at the spreadsheet to see the difference.
GPT-5.6 Luna's sharpest advantage is in knowledge, where it averages 92.3 against 56.6. The single biggest benchmark swing on the page is Terminal-Bench 2.0, 84.7% to 52.5%. Qwen3.5 397B does hit back in mathematics, so the answer changes if that is the part of the workload you care about most.
GPT-5.6 Luna is also the more expensive model on tokens at $1.00 input / $6.00 output per 1M tokens, versus $0.60 input / $3.60 output per 1M tokens for Qwen3.5 397B. GPT-5.6 Luna 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. GPT-5.6 Luna 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 | GPT-5.6 Luna | Δ | Qwen3.5 397B |
|---|---|---|---|
| Knowledge | GPT-5.6 Luna92.3 | Margin← 35.7 | Qwen3.5 397B56.6 |
| Agentic | GPT-5.6 Luna84.1 | Margin← 27.6 | Qwen3.5 397B56.5 |
| Math | GPT-5.6 Luna73.6 | Margin→ 17.0 | Qwen3.5 397B90.6 |
| Coding | GPT-5.6 Luna62.7 | Margin→ 3.8 | Qwen3.5 397B66.5 |
| Multimodal | GPT-5.6 Luna78.4 | Margin→ 1.2 | Qwen3.5 397B79.6 |
| Reasoning | GPT-5.6 LunaNot measured | MarginNo overlap | Qwen3.5 397B63.2 |
| Multilingual | GPT-5.6 LunaNot measured | MarginNo overlap | Qwen3.5 397B84.7 |
| Inst. Following | GPT-5.6 LunaNot measured | MarginNo overlap | Qwen3.5 397B92.6 |
Decisive benchmark drivers
The largest measured benchmark gaps in this matchup, with exact reported values.
More
- Source ↗
Terminal-Bench 2.0
AgenticA 84.7%B 52.5%Winner: GPT-5.6 LunaΔ 32.2Terminal-Bench 2.0: GPT-5.6 Luna scored 84.7%; Qwen3.5 397B scored 52.5%. GPT-5.6 Luna wins this benchmark. - Source ↗
BrowseComp
AgenticA 83.3%B 62%Winner: GPT-5.6 LunaΔ 21.3BrowseComp: GPT-5.6 Luna scored 83.3%; Qwen3.5 397B scored 62%. GPT-5.6 Luna wins this benchmark. - Source ↗
SWE-bench Pro
CodingA 62.7%B 50.9%Winner: GPT-5.6 LunaΔ 11.8SWE-bench Pro: GPT-5.6 Luna scored 62.7%; Qwen3.5 397B scored 50.9%. GPT-5.6 Luna wins this benchmark. - Source ↗
GPQA
KnowledgeA 92.3%B 88.4%Winner: GPT-5.6 LunaΔ 3.9GPQA: GPT-5.6 Luna scored 92.3%; Qwen3.5 397B scored 88.4%. GPT-5.6 Luna wins this benchmark. - Source ↗
MMMU-Pro
MultimodalA 78.4%B 79%Winner: Qwen3.5 397BΔ 0.6MMMU-Pro: GPT-5.6 Luna scored 78.4%; Qwen3.5 397B scored 79%. Qwen3.5 397B wins this benchmark.
Operational comparison
Runtime and commercial metrics are compared only when both models have a complete sourced value.
| Metric | GPT-5.6 Luna | Qwen3.5 397B | Comparison |
|---|---|---|---|
| Input / output priceUSD per 1M tokens | GPT-5.6 Luna$1 input / $6 output | Qwen3.5 397B$0.6 input / $3.6 output | Qwen3.5 397B has the lower combined listed price. |
| Generation speedtokens per second | GPT-5.6 LunaNot available | Qwen3.5 397B96 tok/s | A complete speed comparison is not available. |
| First-answer latencyseconds to first token | GPT-5.6 LunaNot available | Qwen3.5 397B2.44 s | A complete latency comparison is not available. |
| Context windowmaximum listed tokens | GPT-5.6 Luna1M | Qwen3.5 397B128K | GPT-5.6 Luna lists the larger context window. |
Benchmark Deep Dive
AgenticGPT-5.6 Luna wins26 benchmarks
| Benchmark | GPT-5.6 Luna | Qwen3.5 397B | Result |
|---|---|---|---|
| Terminal-Bench 2.0Source | 84.7% | 52.5% | GPT-5.6 Luna leads |
| BrowseCompSource | 83.3% | 62% | GPT-5.6 Luna leads |
| OSWorld 2.0Source | 45.6% | — | Not comparable |
| CyberGymSource | 77.9% | — | Not comparable |
| ExploitGymSource | 12.4% | — | Not comparable |
| ToolathlonSource | 53.4% | 36.3% | GPT-5.6 Luna leads |
| AA Agentic IndexSource | 45.6% | 19.9% | GPT-5.6 Luna leads |
| GDPval-AASource | 54.2% | 23.1% | GPT-5.6 Luna leads |
| GDPval-AASource | 1584 | 962 | GPT-5.6 Luna leads |
| AA Harvey LABSource | 87.9% | — | Not comparable |
| AA ITBenchSource | 40.3% | — | Not comparable |
| AA Tau3 BankingSource | 27.2% | — | Not comparable |
| AA AutomationBenchSource | 42.2% | — | Not comparable |
| aaTerminalBench21Source | 80.9% | — | Not comparable |
| APEX-Agents-AASource | 35.8% | 15.3% | GPT-5.6 Luna leads |
| Claw-EvalSource | — | 56.8% | Not comparable |
| QwenClawBenchSource | — | 51.8% | Not comparable |
| τ³-bench resultsSource | — | 68.4% | Not comparable |
| VITA-BenchSource | — | 43.7% | Not comparable |
| DeepPlanningSource | — | 37.6% | Not comparable |
| MCP AtlasSource | — | 46.1% | Not comparable |
| MCP-TasksSource | — | 74.2% | Not comparable |
| WideResearchSource | — | 74.0% | Not comparable |
| τ²-bench resultsSource | — | 95.6% | Not comparable |
| Gert LabsSource | — | 46.76% | Not comparable |
| ResearchClawBenchSource | — | 14.2% | Not comparable |
CodingQwen3.5 397B wins9 benchmarks
| Benchmark | GPT-5.6 Luna | Qwen3.5 397B | Result |
|---|---|---|---|
| SWE-bench ProSource | 62.7% | 50.9% | GPT-5.6 Luna leads |
| Terminal-Bench 2.0Source | 84.7% | — | Not comparable |
| deepSweSource | 67.2% | — | Not comparable |
| FrontierCode 1.1 ExtendedSource | 55.1% | — | Not comparable |
| cursorBench32Source | 61.1% | — | Not comparable |
| AA Coding IndexSource | 71.5% | 48.2% | GPT-5.6 Luna leads |
| AA-SciCodeSource | 52.5% | 42.0% | GPT-5.6 Luna leads |
| SWE-bench VerifiedSource | — | 76.2% | Not comparable |
| LiveCodeBench v6Source | — | 83.6% | Not comparable |
Reasoning5 benchmarks
KnowledgeGPT-5.6 Luna wins15 benchmarks
| Benchmark | GPT-5.6 Luna | Qwen3.5 397B | Result |
|---|---|---|---|
| GPQASource | 92.3% | 88.4% | GPT-5.6 Luna leads |
| GPQA-DSource | 92.3% | — | Not comparable |
| HealthBench ProfessionalSource | 55.7% | — | Not comparable |
| HealthBench HardSource | 32.0% | — | Not comparable |
| Artificial Analysis Intelligence IndexSource | 51.2% | 33.7% | GPT-5.6 Luna leads |
| AA-GPQA DiamondSource | 91.1% | 89.3% | GPT-5.6 Luna leads |
| AA-HLESource | 37.2% | 27.3% | GPT-5.6 Luna leads |
| AA-Omniscience IndexSource | -11.2% | -29.8% | GPT-5.6 Luna leads |
| AA-Omniscience AccuracySource | 41.5% | 31.4% | GPT-5.6 Luna leads |
| AA-Omniscience Hallucination RateSource | 90.1% | 89.1% | Qwen3.5 397B leads |
| SuperGPQASource | — | 70.4% | Not comparable |
| MMLU-ProSource | — | 87.8% | Not comparable |
| MMLU-ReduxSource | — | 94.9% | Not comparable |
| C-EvalSource | — | 93% | Not comparable |
| HLESource | — | 28.7% | Not comparable |
MathQwen3.5 397B wins8 benchmarks
| Benchmark | GPT-5.6 Luna | Qwen3.5 397B | Result |
|---|---|---|---|
| FrontierMath (legacy)Source | 78.6% | — | Not comparable |
| FrontierMath v2 (Tiers 1-3)Source | 78.600% | — | Not comparable |
| FrontierMath v2 (Tier 4)Source | 58.500% | — | Not comparable |
| AIME26Source | — | 93.3% | Not comparable |
| HMMT Feb 2025Source | — | 94.8% | Not comparable |
| HMMT Nov 2025Source | — | 92.7% | Not comparable |
| HMMT Feb 2026Source | — | 87.9% | Not comparable |
| MMAnswerBenchSource | — | 80.9% | Not comparable |
Multilingual2 benchmarks
MultimodalQwen3.5 397B wins8 benchmarks
| Benchmark | GPT-5.6 Luna | Qwen3.5 397B | Result |
|---|---|---|---|
| MMMU-ProSource | 78.4% | 79% | Qwen3.5 397B leads |
| MMMU-Pro w/ PythonSource | 79.5% | — | Not comparable |
| AA-MMMU-ProSource | 78.6% | 77.3% | GPT-5.6 Luna leads |
| MathVisionSource | — | 88.6% | Not comparable |
| CharXivSource | — | 80.8% | Not comparable |
| VideoMMMUSource | — | 84.7% | Not comparable |
| ScreenSpot ProSource | — | 65.6% | Not comparable |
| V*Source | — | 95.8% | Not comparable |
Frequently Asked Questions (6)
Which is better, GPT-5.6 Luna or Qwen3.5 397B?
GPT-5.6 Luna is ahead on BenchLM's BenchAlign leaderboard, 67.17 to 57.01. The biggest single separator in this matchup is Terminal-Bench 2.0, where the scores are 84.7% and 52.5%.
Which is better for knowledge tasks, GPT-5.6 Luna or Qwen3.5 397B?
GPT-5.6 Luna has the edge for knowledge tasks in this comparison, averaging 92.3 versus 56.6. Inside this category, AA-Omniscience Index is the benchmark that creates the most daylight between them.
Which is better for coding, GPT-5.6 Luna or Qwen3.5 397B?
Qwen3.5 397B has the edge for coding in this comparison, averaging 66.5 versus 62.7. Inside this category, AA Coding Index is the benchmark that creates the most daylight between them.
Which is better for math, GPT-5.6 Luna or Qwen3.5 397B?
Qwen3.5 397B has the edge for math in this comparison, averaging 90.6 versus 73.6. GPT-5.6 Luna stays close enough that the answer can still flip depending on your workload.
Which is better for agentic tasks, GPT-5.6 Luna or Qwen3.5 397B?
GPT-5.6 Luna has the edge for agentic tasks in this comparison, averaging 84.1 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, GPT-5.6 Luna or Qwen3.5 397B?
Qwen3.5 397B has the edge for multimodal and grounded tasks in this comparison, averaging 79.6 versus 78.4. Inside this category, AA-MMMU-Pro is the benchmark that creates the most daylight between them.
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.