BenchLeader
NVIDIAOpen weightsReasoning modelNewReleased 15 Oct 2026Compare with another model

Nemotron 3.5 Super VL 120B A12B

Nemotron 3.5 Super VL 120B A12B is a NVIDIA open-weights reasoning model, released 15 Oct 2026. Its best configuration ranks #280 of 777 on the BenchLeader Index at 53.7 ±4.4, in the upper half. The ± is the point: 293 other configurations score within that range, so they and this one cannot be told apart on quality alone — price and speed are what separate them. It scores highest in human preference (59) and lowest in coding (49). At $0.788 per million tokens blended it is mid-priced. Last measured 20 Oct 2026.

Blended price
$0.788/M
$0.570 in · $1.44 out
Output speed
–
First answer
–
Context
1.0M
Full answer
–
Cost per run
–
Released
15 Oct 2026
How it scores by categoryDashed line = average model (50). One step of 15 = one standard deviation.
  1. Overall index54
  2. Reasoning59
  3. Coding49
  4. Maths58
  5. Human preference59
  6. Multimodal52

Benchmark results

See all LLM benchmarks. One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.

Reasoning

Benchmarknot statedSource
Arena Hard Prompts1433#133Arena (LMArena)

Coding

Benchmarknot statedSource
Arena Coding1456#139Arena (LMArena)
Arena WebDev1404#86Arena (LMArena)

Maths

Benchmarknot statedSource
Arena Maths1407#139Arena (LMArena)

Instruction following

Benchmarknot statedSource
Arena Instruction Followingnot in index1411#122Arena (LMArena)

Human preference

Benchmarknot statedSource
Arena Text1414#135Arena (LMArena)
Arena Creative Writingnot in index1389#123Arena (LMArena)
Arena Multi-turnnot in index1429#115Arena (LMArena)
Arena Longer Queriesnot in index1416#138Arena (LMArena)

Multimodal

Benchmarknot statedSource
Arena Vision1205#89Arena (LMArena)

What a task costs

Estimates from list price, output speed and time to first answer for the best configuration. “With caching” assumes three-quarters of the input is served from the prompt cache at $0.114 per 1M. Reasoning tokens are not modelled.

WorkloadTokens in / outCostWith cachingTime
Chat reply400 / 300$0.0007$0.0005–
Summarize a 30-page report12,000 / 600$0.0077$0.0036–
Code edit6,000 / 1,500$0.0056$0.0035–
Agentic coding session60,000 / 4,000$0.040$0.019–
Structured extraction2,000 / 200$0.0014$0.0007–

More from NVIDIA

New this month

See also

Data as of 21 Oct 2026. Compare with another model.

Cite as: BenchLeader, “Nemotron 3.5 Super VL 120B A12B: benchmarks, pricing, speed and rank”, https://www.benchleader.com/models/nemotron-3-5-super-vl-120b-a12b, data as of 21 Oct 2026.