Llama Nemotron Super 49B v1.5 vs Nemotron 3.5 Super VL
Humanity's Last Exam Humanity's Last Exam: Very hard expert questions across many fields, written to stump today’s models
Llama Nemotron Super 49B v1.54.3
Nemotron 3.5 Super VL30.7
Terminal-Bench 4.0 Terminal-Bench 4.0 · source setting 2: Agents use a terminal to complete executable coding, system administration and research tasks in the 4.0 suite
Llama Nemotron Super 49B v1.5—
Nemotron 3.5 Super VL0
AutomationBench-AA AutomationBench-AA: Agents orchestrate SaaS workflows through APIs; objectives receive credit only when the task has no guardrail violation
Llama Nemotron Super 49B v1.5—
Nemotron 3.5 Super VL14.8
Arena Text · creative writing Arena Text · creative writing: Human pairwise preferences in the stated text vote pool
Llama Nemotron Super 49B v1.51,305
Nemotron 3.5 Super VL—
| Benchmark | Llama Nemotron Super 49B v1.5 | Nemotron 3.5 Super VL |
|---|---|---|
| Humanity's Last Exam | 4.3 | 30.7 |
| Terminal-Bench 4.0 | — | 0 |
| AutomationBench-AA | — | 14.8 |
| Arena Text · creative writing | 1,305 | — |
Specs
| Spec | Llama Nemotron Super 49B v1.5 | Nemotron 3.5 Super VL |
|---|---|---|
| Input, per 1M | $0.40 | $0.10 |
| Output, per 1M | $0.40 | $0.40 |
| Cost per task? | — | $0.19 |
| Speed, tok/s | — | 413 |
| Context | 128K | 1M |
| Weights | Open | Open |
| Released | Jul 2025 | Oct 2026 |
| Memory (4-bit) | 30 GB | 75 GB |
| Licence | nvidia-open-model-license | openmdw-1.1 |
| Commercial use | Restricted | Restricted |
| See images | — | ✓ |
| Read PDFs | — | — |
| Hear audio | — | — |
| Watch video | — | ✓ |
| Call tools | ✓ | ✓ |
| Force a tool call | ? | Partial |
| Return strict JSON | Partial | ✓ |
| Search the web | ? | ? |
| Run code | ? | ? |
| Use a computer | ? | ? |
What breaks when you switch
Llama Nemotron Super 49B v1.5 → Nemotron 3.5 Super VL
Memory (4-bit): 30 GB → 75 GB
Nemotron 3.5 Super VL → Llama Nemotron Super 49B v1.5