Nemotron 3.5 Super VL 120B A12B vs Qwen3.5-9B
Full profiles: Nemotron 3.5 Super VL 120B A12B (NVIDIA) and Qwen3.5-9B (Alibaba): every benchmark, price and rank.
Updated · Data from 4 sources
As of 9 Nov 2026, Nemotron 3.5 Super VL 120B A12B and Qwen3.5-9B are neck and neck on quality (50.3 vs 50.4 on the BenchLeader Index, inside the ±13.2 margin), and Qwen3.5-9B is 4.1× cheaper at $0.190 vs $0.788 per 1M tokens blended. Choose Nemotron 3.5 Super VL 120B A12B for reasoning (59 vs 35) and multimodal (54 vs 51); choose Qwen3.5-9B for agents & tools (41 vs 24).
Key facts (as of 9 Nov 2026)
- Nemotron 3.5 Super VL 120B A12B scores 50.3 on the BenchLeader Index and Qwen3.5-9B scores 50.4 (BenchLeader, 9 Nov 2026).
- Nemotron 3.5 Super VL 120B A12B costs $0.788 per 1M tokens blended (list price, 9 Nov 2026).
- Qwen3.5-9B costs $0.190 per 1M tokens blended (list price, 9 Nov 2026).
Nemotron 3.5 Super VL 120B A12B benchmarks NVIDIA
Qwen3.5-9B benchmarks thinking AlibabaAdd a third modelWhich to pick for…
The pick's numbers come first. Category scores on the BenchLeader scale (50 is the average model, 15 points one standard deviation). High confidence means a lead of half a standard deviation or more; under one point is called either.
| Use case | Pick | Why | Confidence |
|---|---|---|---|
| Agents & tool use | Qwen3.5-9B | 41 vs 24 category score | High |
| Reasoning | Nemotron 3.5 Super VL 120B A12B | 59 vs 35 category score | High |
| Images and other media | Nemotron 3.5 Super VL 120B A12B | 54 vs 51 category score | Medium |
| Chat quality (human preference) | – | Only Nemotron 3.5 Super VL 120B A12B has an Arena rating (59 category score) | Not comparable |
| Lowest cost | Qwen3.5-9B | $0.155 vs $0.788 per 1M blended; agentic coding session $0.009 vs $0.040 | – |
| High volume on a budget | Neither flagship | A smaller model is usually the better fit: Cheapest good LLMs, Fastest LLMs | – |
Key differences
- Nemotron 3.5 Super VL 120B A12B and Qwen3.5-9B (thinking) are neck and neck on quality (50.3 vs 50.4, within the ±13.2 margin).
- Nemotron 3.5 Super VL 120B A12B is stronger in coding, human preference, maths, multimodal, reasoning.
- Qwen3.5-9B (thinking) is stronger in agents & tools, instruction following, long context.
- Qwen3.5-9B (thinking) is 4.1× cheaper ($0.190 vs $0.788 per 1M blended).
At a glance
| Spec | Nemotron 3.5 Super VL 120B A12B | Qwen3.5-9B |
|---|---|---|
| Developer | NVIDIA | Alibaba |
| Released | 15 Oct 2026 | 24 Feb 2026 |
| Context window (tokens) | 1,048,576 | 262,144 |
| Compared at | Nemotron 3.5 Super VL 120B A12B | Qwen3.5-9B (thinking) |
| Open weights | Yes | Yes |
| Price in / out / cached input per 1M | $0.570 / $1.44 / $0.114 | $0.140 / $0.200 / $0.060 |
Category by category
Category scores on the BenchLeader scale: 50 is the average evaluated model, 15 points is one standard deviation. A gap under one point is called level.
| Category | Nemotron 3.5 Super VL 120B A12B | Qwen3.5-9B (thinking) | Leader |
|---|---|---|---|
| Reasoning | 58.9 | 35.0 | Nemotron 3.5 Super VL 120B A12B |
| Coding | 48.1 | – | not comparable |
| Agents & tools | 23.9 | 41.3 | Qwen3.5-9B |
| Maths | 59.2 | – | not comparable |
| Instruction following | – | 66.0 | not comparable |
| Human preference | 59.1 | – | not comparable |
| Multimodal | 53.5 | 51.4 | Nemotron 3.5 Super VL 120B A12B |
| Long context | – | 59.5 | not comparable |
On one chart
Each domain on the index's own scale, where 50 is an average model. Click a domain to see the benchmarks behind it. Pick up to 5 models from the list.
Not drawn, measured for fewer than two of these models: Coding, Maths, Knowledge, Instruction following, Long context, Human preference.
Where sources disagree
Across the 0 benchmarks both have, Nemotron 3.5 Super VL 120B A12B leads 0, Qwen3.5-9B leads 0.
| Overall index | Nemotron 3.5 Super VL 120B A12B | Qwen3.5-9B (thinking) | Leader |
|---|---|---|---|
| BenchLeader Index | 50.3 | 50.4 | neck and neck |
Each publisher runs its own tasks, harnesses and effort settings, so their overall verdicts can differ; BenchLeader reconciles them on one scale. † measured at another effort setting of the same model (the publisher did not test the one compared here).
Price, speed and latency
List prices blended at three input tokens per output token. Per-task prices use the token counts on the methodology page. Cost, tokens and time per benchmark task are measured: Artificial Analysis’ averages over its whole Intelligence Index run.
| Measure | Nemotron 3.5 Super VL 120B A12B | Qwen3.5-9B (thinking) |
|---|---|---|
| Blended price per 1M tokens | $0.788 | $0.155 |
| Input / output per 1M | $0.570 / $1.44 | $0.140 / $0.200 |
| Cached input per 1M | $0.114 | $0.060 |
| Price per task: chat reply | $0.0007 | $0.0001 |
| Price per task: summarize a 30-page report | $0.0077 | $0.0018 |
| Price per task: code edit | $0.0056 | $0.0011 |
| Price per task: agentic coding session | $0.040 | $0.0092 |
| Price per task: structured extraction | $0.0014 | $0.0003 |
| Agentic coding session, input mostly cached | $0.019 | $0.0056 |
| Output speed | – | 86 tok/s |
| Time to first answer | – | 40.2 s |
| Time to a complete answer | – | 102.8 s |
Each model is compared at its best-ranked reasoning-effort setting. Nemotron 3.5 Super VL 120B A12B scores 50.3. Qwen3.5-9B (thinking) scores 50.4 and takes 40 s to its first answer token; Qwen3.5-9B (no reasoning) scores 49.2 in 1 s, Qwen3.5-9B scores 39.9 in 0 s. Higher effort means more thinking before the answer: better scores, longer waits and more tokens billed. How reasoning effort changes scores and cost.
Every benchmark
50 benchmarks, grouped by category; each name links to the board and its leaders. † measured at another effort setting of the same model, shown for reference and not counted as a win.
| Metric | Publisher | Nemotron 3.5 Super VL 120B A12B | Qwen3.5-9B (thinking) |
|---|---|---|---|
| BenchLeader Index | BenchLeader | 50.3 | 50.4 |
| Agents & tools score | BenchLeader | 23.9 | 41.3 |
| Coding score | BenchLeader | 48.1 | – |
| Human preference score | BenchLeader | 59.1 | – |
| Maths score | BenchLeader | 59.2 | – |
| Multimodal score | BenchLeader | 53.5 | 51.4 |
| Reasoning score | BenchLeader | 58.9 | 35.0 |
| Instruction following score | BenchLeader | – | 66.0 |
| Long context score | BenchLeader | – | 59.5 |
| Blended price $/M | BenchLeader | $0.788 | $0.155 |
| Output speed | BenchLeader | – | 86 tok/s |
| Time to first answer | BenchLeader | – | 40.2 s |
| Context window | BenchLeader | 1,048,576 tokens | 262,144 tokens |
| Reasoning | |||
| GPQA Diamond | Epoch AI Benchmarking Hub | – | 79.0%† |
| Arena Hard Prompts | Arena (LMArena) | 1432 | – |
| GPQA Diamond (AA) | Artificial Analysis | – | 80.6% |
| Humanity's Last Exam (AA) | Artificial Analysis | – | 14.9% |
| HLE-Diamond (AA) | Artificial Analysis | – | 4.8% |
| CritPt | Artificial Analysis | – | 0.3% |
| LMCA | Epoch AI Benchmarking Hub | – | 24.5%† |
| DTBench | Epoch AI Benchmarking Hub | – | 71.2%† |
| Coding | |||
| SciCode | SciCode | – | 29.5%† |
| Arena Coding | Arena (LMArena) | 1460 | – |
| Arena WebDev | Arena (LMArena) | 1399 | – |
| Code Migration v1.1 | Vals AI | 16.4%† | – |
| Vibe Code Bench v1.1 | Vals AI | 3.7%† | – |
| Terminal-Bench 4.0 (AA) | Artificial Analysis | – | 0.5% |
| Terminal-Bench 2.1 (AA) | Artificial Analysis | – | 29.2% |
| Agents & tools | |||
| Terminal-Bench | Terminal-Bench | – | 9.2%† |
| Arena Agent | Arena (LMArena) | -16.3 | – |
| Terminal-Bench Hard | Artificial Analysis | – | 24.2% |
| τ²-Bench Telecom (AA) | Artificial Analysis | – | 86.8% |
| GDPval-AA v2.1, normalised score | Artificial Analysis | – | 0.0% |
| GDPval-AA v2.1 | Artificial Analysis | – | 427 |
| τ³-Banking (AA) | Artificial Analysis | – | 7.0% |
| Finance Agent v2.1 | Vals AI | 47.7%† | – |
| Harvey's Legal Agent Benchmark v1.1 | Vals AI | 3.8%† | – |
| Legal Research Bench | Vals AI | 10.1%† | – |
Show all 50 benchmarks
| Metric | Publisher | Nemotron 3.5 Super VL 120B A12B | Qwen3.5-9B (thinking) |
|---|---|---|---|
| Agents & tools | |||
| Tax Agent Bench | Vals AI | 11.6%† | – |
| Terminal-Bench 4.0 (Vals) | Vals AI | 0.0%† | – |
| AutomationBench | Artificial Analysis | – | 1.1% |
| GDP.pdf (AA) | Artificial Analysis | – | 6.6% |
| AA-Briefcase v1.1 | Artificial Analysis | – | 341 |
| Maths | |||
| OTIS Mock AIME | Epoch AI Benchmarking Hub | – | 42.8%† |
| AIME 2026 | MathArena | – | 92.5%† |
| HMMT February 2026 | MathArena | – | 71.2%† |
| MathArena Apex | MathArena | – | 0.5%† |
| Arena Maths | Arena (LMArena) | 1417 | – |
| Knowledge | |||
| Excel Modeling Benchmark v1.1 | Vals AI | 52.8%† | – |
| Instruction following | |||
| IFBench | Artificial Analysis | – | 66.7% |
| Arena Instruction Following | Arena (LMArena) | 1411 | – |
| Human preference | |||
| Arena Text | Arena (LMArena) | 1415 | – |
| Arena Creative Writing | Arena (LMArena) | 1388 | – |
| Arena Multi-turn | Arena (LMArena) | 1425 | – |
| Arena Longer Queries | Arena (LMArena) | 1412 | – |
| Multimodal | |||
| Arena Vision | Arena (LMArena) | 1214 | – |
| MMMU-Pro | Artificial Analysis | – | 69.3% |
| Long context | |||
| AA-LCR v1.1 | Artificial Analysis | – | 70.0% |
| Composite | |||
| Epoch Capabilities Index | Epoch AI Benchmarking Hub | – | 139.3† |
| AA Intelligence Index v4.3.2 | Artificial Analysis | – | 11.2 |
| AA Progress Index v1 | Artificial Analysis | – | 239.6 |
| Vals Index | Vals AI | 29.8† | – |
| Vals Finance Index | Vals AI | 41.0%† | – |
Data as of 2026-11-09. Best configuration of each model; every score links to its source on the model pages.
What changed
- Nemotron 3.5 Super VL 120B A12B: 1 score revision.
- Nemotron 3.5 Super VL 120B A12B: 1 score revision.
- Nemotron 3.5 Super VL 120B A12B: 1 score revision.
- Nemotron 3.5 Super VL 120B A12B: 1 score revision.
- Qwen3.5-9B (thinking): 1 new result (AA-Briefcase v1.1).
- Qwen3.5-9B (thinking): BenchLeader Index 51.0 → 50.5.
- Qwen3.5-9B (thinking): 2 new results (AA Progress Index v1, HLE-Diamond (AA)).
- Nemotron 3.5 Super VL 120B A12B: BenchLeader Index 53.6 → 50.4.
- Nemotron 3.5 Super VL 120B A12B: 1 new result (Arena Agent).
- Nemotron 3.5 Super VL 120B A12B: 1 score revision.
Related comparisons
- Nemotron 3.5 Super VL 120B A12B vs Inkling
- Nemotron 3.5 Super VL 120B A12B vs Qwen3.5-35B-A3B
- Qwen3.5-9B vs Agnes 2.5 Pro Theta
Rankings: Best LLM overall · Best LLM for agents · Best cheap LLMs
Nemotron 3.5 Super VL 120B A12B vs Qwen3.5-9B: questions
- Is Nemotron 3.5 Super VL 120B A12B better than Qwen3.5-9B?
- Nemotron 3.5 Super VL 120B A12B and Qwen3.5-9B (thinking) are neck and neck on quality (50.3 vs 50.4, within the ±13.2 margin). The BenchLeader Index combines every independent quality benchmark; the two are within the margin of error as of 9 Nov 2026, so check the category scores for your use.
- Nemotron 3.5 Super VL 120B A12B vs Qwen3.5-9B: which is cheaper, and will that change?
- Qwen3.5-9B is cheaper today: $0.190 against $0.788 per million tokens, blended at three input tokens per output token. No price change has been announced for either.
- Is Nemotron 3.5 Super VL 120B A12B or Qwen3.5-9B better for agents and tool use?
- Qwen3.5-9B scores higher in agentic tasks (41 vs 24 on the category index, where 50 is average).
- Which is better for long documents?
- Nemotron 3.5 Super VL 120B A12B accepts more context: 1,048,576 against 262,144 tokens for Qwen3.5-9B.
- When was Nemotron 3.5 Super VL 120B A12B released?
- Nemotron 3.5 Super VL 120B A12B was released on 15 Oct 2026.