Mistral Large 3 vs Nemotron 3.5 Super VL
Full profiles: Mistral Large 3 (Mistral AI) and Nemotron 3.5 Super VL (NVIDIA): every benchmark, price and rank.
Updated · Data from 9 sources
As of 16 Nov 2026, Mistral Large 3 and Nemotron 3.5 Super VL are neck and neck on quality (45.1 vs 45.8 on the BenchLeader Index, inside the ±8.9 margin), and Nemotron 3.5 Super VL is 2.7× cheaper at $0.275 vs $0.750 per 1M tokens blended. Choose Mistral Large 3 for knowledge (51 vs 43), coding (37 vs 34), and reasoning (38 vs 37) and fast first answers (2.6 s vs 13.7 s); choose Nemotron 3.5 Super VL for long context (59 vs 42), where it also streams 3×+ faster.
Key facts (as of 16 Nov 2026)
- Mistral Large 3 scores 45.1 on the BenchLeader Index and Nemotron 3.5 Super VL scores 45.8 (BenchLeader, 16 Nov 2026).
- Mistral Large 3 costs $0.750 per 1M tokens blended (list price, 16 Nov 2026).
- Nemotron 3.5 Super VL costs $0.275 per 1M tokens blended (list price, 16 Nov 2026).
- Nemotron 3.5 Super VL streams 3×+ faster than Mistral Large 3 (Artificial Analysis, 7-day median, 16 Nov 2026).
- Of the 14 benchmarks both models have, Nemotron 3.5 Super VL leads 13 (BenchLeader, 16 Nov 2026).
Which 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 |
|---|---|---|---|
| Coding | Mistral Large 3 | 37 vs 34 category score; SciCode (AA) 36.6% vs 47.8% | Medium |
| Reasoning | Mistral Large 3 | 38 vs 37 category score; HLE-Diamond (AA) 4.5% vs 8.8% | Medium |
| Knowledge | Mistral Large 3 | 51 vs 43 category score | High |
| Long documents | Nemotron 3.5 Super VL | 59 vs 42 category score; AA-LCR v1.1 69.0% vs 36.0% | High |
| Chat quality (human preference) | – | Only Mistral Large 3 has an Arena rating (59 category score) | Not comparable |
| Lowest cost | Nemotron 3.5 Super VL | $0.275 vs $0.750 per 1M blended; agentic coding session $0.014 vs $0.036 | – |
| Fastest first answer | Mistral Large 3 | 2.6 s (Mistral Large 3) vs 13.7 s (Nemotron 3.5 Super VL) to the first answer token | – |
| High volume on a budget | Neither flagship | A smaller model is usually the better fit: Cheapest good LLMs, Fastest LLMs | – |
Key differences
- Mistral Large 3 and Nemotron 3.5 Super VL are neck and neck on quality (45.1 vs 45.8, within the ±8.9 margin).
- Mistral Large 3 is stronger in coding, knowledge, reasoning.
- Nemotron 3.5 Super VL is stronger in long context.
- Nemotron 3.5 Super VL is 2.7× cheaper ($0.275 vs $0.750 per 1M blended).
- Nemotron 3.5 Super VL streams 3.2× faster (271 vs 84 tokens per second).
At a glance
| Spec | Mistral Large 3 | Nemotron 3.5 Super VL |
|---|---|---|
| Developer | Mistral AI | NVIDIA |
| Released | 2 Dec 2025 | 15 Oct 2026 |
| Context window (tokens) | 262,144 | 1,048,576 |
| Compared at | Mistral Large 3 | Nemotron 3.5 Super VL |
| Open weights | Yes | Yes |
| Price in / out / cached input per 1M | $0.500 / $1.50 / $0.050 | $0.200 / $0.500 / $0.020 |
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 | Mistral Large 3 | Nemotron 3.5 Super VL | Leader |
|---|---|---|---|
| Reasoning | 38.2 | 36.5 | Mistral Large 3 |
| Coding | 37.1 | 34.4 | Mistral Large 3 |
| Agents & tools | 37.9 | – | not comparable |
| Maths | 49.6 | – | not comparable |
| Knowledge | 51.1 | 43.2 | Mistral Large 3 |
| Instruction following | 39.4 | – | not comparable |
| Human preference | 59.0 | – | not comparable |
| Multimodal | 46.0 | – | not comparable |
| Long context | 41.7 | 58.9 | Nemotron 3.5 Super VL |
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: Agents & tools, Maths, Instruction following, Multimodal, Human preference.
Where sources disagree
Across the 14 benchmarks both have, Mistral Large 3 leads 0, Nemotron 3.5 Super VL leads 13 and 1 is a tie.
| Overall index | Mistral Large 3 | Nemotron 3.5 Super VL | Leader |
|---|---|---|---|
| BenchLeader Index | 45.1 | 45.8 | 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 | Mistral Large 3 | Nemotron 3.5 Super VL |
|---|---|---|
| Blended price per 1M tokens | $0.750 | $0.275 |
| Input / output per 1M | $0.500 / $1.50 | $0.200 / $0.500 |
| Cached input per 1M | $0.050 | $0.020 |
| Price per task: chat reply | $0.0007 | $0.0002 |
| Price per task: summarize a 30-page report | $0.0069 | $0.0027 |
| Price per task: code edit | $0.0053 | $0.0020 |
| Price per task: agentic coding session | $0.036 | $0.014 |
| Price per task: structured extraction | $0.0013 | $0.0005 |
| Agentic coding session, input mostly cached | $0.016 | $0.0059 |
| Cost per benchmark task (AA Index) | – | $0.557 |
| Output tokens per task (AA Index) | – | 81k |
| Of which reasoning | – | 56k |
| Time per task (AA Index) | – | 5 min |
| Output speed | 84 tok/s | 271 tok/s |
| Time to first answer | 2.6 s | 13.7 s |
| Time to a complete answer, last measured 13 Nov 2026 | 34.1 s | 36.4 s |
Each model is compared at its best-ranked reasoning-effort setting. Mistral Large 3 scores 45.1 and takes 3 s to its first answer token. Nemotron 3.5 Super VL scores 45.8 and takes 14 s to its first answer token. 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
74 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 | Mistral Large 3 | Nemotron 3.5 Super VL |
|---|---|---|---|
| BenchLeader Index | BenchLeader | 45.1 | 45.8 |
| Agents & tools score | BenchLeader | 37.9 | – |
| Coding score | BenchLeader | 37.1 | 34.4 |
| Human preference score | BenchLeader | 59.0 | – |
| Instruction following score | BenchLeader | 39.4 | – |
| Knowledge score | BenchLeader | 51.1 | 43.2 |
| Long context score | BenchLeader | 41.7 | 58.9 |
| Maths score | BenchLeader | 49.6 | – |
| Multimodal score | BenchLeader | 46.0 | – |
| Reasoning score | BenchLeader | 38.2 | 36.5 |
| Blended price $/M | BenchLeader | $0.750 | $0.275 |
| Output speed | BenchLeader | 84 tok/s | 271 tok/s |
| Time to first answer | BenchLeader | 2.6 s | 13.7 s |
| Context window | BenchLeader | 262,144 tokens | 1,048,576 tokens |
| Reasoning | |||
| SimpleBench | SimpleBench | 20.4% | – |
| Arena Hard Prompts | Arena (LMArena) | 1433 | – |
| GPQA Diamond (AA) | Artificial Analysis | 68.0% | – |
| Humanity's Last Exam (AA) | Artificial Analysis | 4.2% | 30.7% |
| HLE-Diamond (AA) | Artificial Analysis | 4.5% | 8.8% |
| GPQA Diamond (Vals) | Vals AI | 68.4% | – |
| Kagi LLM Benchmark | Kagi LLM Benchmark | 50.9% | – |
| ARC-AGI-1 | ARC Prize | – | 2.0%† |
| ARC-AGI-2 | ARC Prize | – | 0.8%† |
| CritPt | Artificial Analysis | 0.0% | 10.6% |
| CritPt v2 hard subset (AA) | Artificial Analysis | – | 13.5% |
| Arena Expert | Arena (LMArena) | 1428 | – |
| LMCA | Epoch AI Benchmarking Hub | 16.7% | – |
| DTBench | Epoch AI Benchmarking Hub | 65.1% | – |
| Coding | |||
| SciCode | SciCode | 36.6% | – |
| Arena Coding | Arena (LMArena) | 1471 | – |
| Arena WebDev | Arena (LMArena) | 1232 | – |
| SciCode (AA) | Artificial Analysis | 36.6% | 47.8% |
| LiveCodeBench | Vals AI | 55.3% | – |
| SWE-bench (Vals) | Vals AI | 41.4% | – |
| SWE Atlas v1.1 | Scale AI SEAL | – | 24.8% |
| SWE Atlas v1.1: Codebase QnA | Scale AI SEAL | – | 26.6% |
| SWE Atlas v1.1: Test Writing | Scale AI SEAL | – | 27.8% |
| SWE Atlas v1.1: Refactoring | Scale AI SEAL | – | 20.0% |
| ALE-Bench | Epoch AI Benchmarking Hub | 264.7 | – |
Show all 74 benchmarks
| Metric | Publisher | Mistral Large 3 | Nemotron 3.5 Super VL |
|---|---|---|---|
| Coding | |||
| Terminal-Bench 4.0 (AA) | Artificial Analysis | 0.0% | 0.0% |
| Terminal-Bench 2.1 (AA) | Artificial Analysis | 12.0% | – |
| Design Arena Website | Design Arena | 1154 | 1160 |
| FrontierCode 1.1 (AA) | Artificial Analysis | – | 25.5% |
| FrontierCode 2.0 (AA) | Artificial Analysis | – | 6.1% |
| LiveCodeBench (AA) | Artificial Analysis | 46.5% | – |
| Agents & tools | |||
| Terminal-Bench | Terminal-Bench | – | 0.3%† |
| Terminal-Bench Hard | Artificial Analysis | 15.9% | – |
| τ²-Bench Telecom (AA) | Artificial Analysis | 24.6% | – |
| GDPval-AA v2.1, normalised score | Artificial Analysis | – | 33.1% |
| GDPval-AA v2.1 | Artificial Analysis | 427 | 1162 |
| τ³-Banking (AA) | Artificial Analysis | 5.8% | – |
| Terminal-Bench 2.0 (Vals) | Vals AI | 9.0% | – |
| AutomationBench | Artificial Analysis | 1.5% | 14.8% |
| GDP.pdf (AA) | Artificial Analysis | 1.2% | 8.0% |
| AA-Briefcase v1.1 | Artificial Analysis | 229 | 999 |
| Terminal-Bench-Science 0.1 (AA) | Artificial Analysis | – | 0.0% |
| Maths | |||
| AIME (Vals) | Vals AI | 42.9% | – |
| MGSM | Vals AI | 85.4% | – |
| Arena Maths | Arena (LMArena) | 1402 | – |
| AIME 2025 (AA) | Artificial Analysis | 38.0% | – |
| Knowledge | |||
| AA-Omniscience v1.1 | Artificial Analysis | – | -2.9 |
| MMLU-Pro | Vals AI | 79.8% | – |
| LegalBench | Vals AI | 79.1% | – |
| CorpFin | Vals AI | 61.0% | – |
| TaxEval | Vals AI | 73.1% | – |
| MedQA | Vals AI | 82.2% | – |
| CaseLaw v2 | Vals AI | 61.4% | – |
| AA-Omniscience v1.1: accuracy | Artificial Analysis | – | 21.2% |
| AA-Omniscience v1.1: non-hallucination | Artificial Analysis | – | 69.5% |
| Global-MMLU-Lite (AA) | Artificial Analysis | 80.3% | – |
| Instruction following | |||
| IFBench | Artificial Analysis | 36.2% | – |
| Arena Instruction Following | Arena (LMArena) | 1403 | – |
| Human preference | |||
| Arena Text | Arena (LMArena) | 1415 | – |
| Arena Creative Writing | Arena (LMArena) | 1375 | – |
| Arena Multi-turn | Arena (LMArena) | 1419 | – |
| Arena Longer Queries | Arena (LMArena) | 1419 | – |
| Multimodal | |||
| Arena Vision | Arena (LMArena) | 1222 | – |
| MMMU-Pro | Artificial Analysis | 55.7% | – |
| MMMU-Pro (Vals) | Vals AI | 66.2% | – |
| MortgageTax | Vals AI | 52.1% | – |
| Humanity's Sixth Sense | Scale AI SEAL | – | 24.7%† |
| Long context | |||
| AA-LCR v1.1 | Artificial Analysis | 36.0% | 69.0% |
| MLCR | Artificial Analysis | 1.1% | 13.3% |
| Composite | |||
| AA Intelligence Index v5.0.1 | Artificial Analysis | – | 13.6 |
| AA Intelligence Index v4.3.2 | Artificial Analysis | 9.3 | 26.5 |
| AA Progress Index v1 | Artificial Analysis | 196.2 | 283.9 |
| Safety & honesty | |||
| DistressBench | Scale AI SEAL | 58.8% | – |
| Metadata | |||
| AA Openness Index | Artificial Analysis | 38.9 | 83.3 |
Data as of 2026-11-16. Best configuration of each model; every score links to its source on the model pages.
What changed
- Mistral Large 3: Mistral Large 3 cached input price moved from $0.070/M to $0.050/M (29% down).
- Nemotron 3.5 Super VL: Nemotron 3.5 Super VL cached input price moved from $0.060/M to $0.020/M (67% down).
- Mistral Large 3: 4 new results (GDPval-AA v2.1, Global-MMLU-Lite (AA), AIME 2025 (AA)…).
- Nemotron 3.5 Super VL: Nemotron 3.5 Super VL input price moved from $0.18/M to $0.20/M (11% up); Nemotron 3.5 Super VL cost per task moved from $0.52/M to $0.56/M (7% up).
- Nemotron 3.5 Super VL: BenchLeader Index 48.9 → 45.8.
- Mistral Large 3: Mistral Large 3 cached input price moved from $0.050/M to $0.070/M (40% up).
- Nemotron 3.5 Super VL: BenchLeader Index 45.7 → 48.9.
- Nemotron 3.5 Super VL: 1 new result (HiL-Bench); 1 score revision.
- Mistral Large 3: BenchLeader Index 45.6 → 45.1.
- Mistral Large 3: 1 new result (SimpleBench).
Related comparisons
- Nemotron 3.5 Super VL vs Command A+
- Nemotron 3.5 Super VL vs Beam 1.0
- Nemotron 3.5 Super VL vs Granite 4.2 30B
Rankings: Best LLM overall · Best LLM for agents · Best cheap LLMs
Mistral Large 3 vs Nemotron 3.5 Super VL: questions
- Is Mistral Large 3 better than Nemotron 3.5 Super VL?
- Mistral Large 3 and Nemotron 3.5 Super VL are neck and neck on quality (45.1 vs 45.8, within the ±8.9 margin). The BenchLeader Index combines every independent quality benchmark; the two are within the margin of error as of 16 Nov 2026, so check the category scores for your use.
- Mistral Large 3 vs Nemotron 3.5 Super VL: which is cheaper, and will that change?
- Nemotron 3.5 Super VL is cheaper today: $0.275 against $0.750 per million tokens, blended at three input tokens per output token. No price change has been announced for either.
- Is Mistral Large 3 or Nemotron 3.5 Super VL better for coding?
- Mistral Large 3 scores higher in coding (37 vs 34 on the category index, where 50 is average).
- Which is faster, Mistral Large 3 or Nemotron 3.5 Super VL?
- Nemotron 3.5 Super VL streams 3×+ faster (271 vs 84 output tokens per second, Artificial Analysis 7-day median).
- Which is better for long documents?
- Nemotron 3.5 Super VL scores higher on long-context benchmarks (59 vs 42). Nemotron 3.5 Super VL accepts more context: 1,048,576 against 262,144 tokens for Mistral Large 3.
More questions (3)
- Which has lower latency?
- Mistral Large 3 answers first: 2.6 s against 13.7 s to the first answer token, as Artificial Analysis measures it. Reasoning effort drives most of this: a higher setting thinks longer before it answers.
- Do independent benchmarks agree on Mistral Large 3 vs Nemotron 3.5 Super VL?
- Across the 14 benchmarks both have, Mistral Large 3 leads 0, Nemotron 3.5 Super VL leads 13 and 1 is a tie.
- When was Nemotron 3.5 Super VL released?
- Nemotron 3.5 Super VL was released on 15 Oct 2026.