Across 1 comparable benchmark, Nemotron 3.5 Super VL has better reported results on 1 and worse on 0.
Limited evidence1 of 1 shared result countedHow we compare
Devstral Small 2 is no longer available through its official API. API prices for this version are not shown.
Price basis: Listed rates vs standard rates. The price section below shows the applicable periods.
| Fact | Devstral Small 2 | Nemotron 3.5 Super VL |
|---|---|---|
| Context window | 256K tokens | 1M tokens |
| Input / 1M | — | $0.150 |
| Output / 1M | — | $0.500 |
| Blended (4:1) | — | $0.126 |
| Status | Retired | Available |
| Released | December 9, 2025 | October 15, 2026 |
Blended is 4:1 input to output with 90% cache hits at listed rates — set your own mix.
No shared benchmark here has a fixed 0–100 scale, so only the counted results are shown. How averages are built
Nemotron 3.5 Super VL ahead on 1 of 1 counted
No fixed-scale average — see the rows in Benchmark results
No shared comparable results for Coding, Abstract reasoning, Mathematics, Science & health, Computer & tool use, Research & search, Long context, Multimodal, Cybersecurity.
Published scores, with matching modes where available. Gaps show Devstral Small 2 relative to Nemotron 3.5 Super VL. Reading the results
| Benchmark | Devstral Small 2 | Nemotron 3.5 Super VL | Gap |
|---|---|---|---|
| τ³-BankingService Workflows · Score | 10.7Normal · tools | 18.9≠Unspecified Reasoning · tools | 0.57×#90 vs #60 of 120 |
Service Workflows · Score
A dash means no result is available.
| Benchmark | Devstral Small 2 | Nemotron 3.5 Super VL |
|---|---|---|
| CritPt | — | 10.9 |
| GPQA Diamond | — | 89.5 |
| HLE | — | 31.8 |
| HLE-Diamond | — | 8.8 |
| DeepSWE | — | 24.6 |
| FrontierCode 2.0 | — | 6.1 |
| SWE-bench Multilingual | — | 67.7 |
| SWE-bench Verified | — | 71.1 |
| Terminal-Bench 2.1 | — | 60.7 |
| Terminal-Bench 4.0 | — | 0.0 |
| Terminal-Bench 4.1 | — | 0.3 |
| Vibe Code Bench v1.1 | — | 3.5 |
| MCP-Atlas | — | 67.7 |
| OSWorld-Verified | — | 51.1 |
| ScreenSpot-Pro (no tools) | — | 43.5 |
| AA Intelligence Index v5.0.1 | — | 13.6 |
| LiveBench 2.0 | — | 56.0 |
| Vals Index v2.2 | — | 29.8 |
| AA-LCR | — | 68.6 |
| LongBench v2 | — | 62.1 |
| OpenAI MRCR v2 (8-needle, 256K–512K) | — | 5.7 |
| BabyVision | — | 29.6 |
| Chartography | — | 29.3 |
| CharXiv RQ | — | 80.8 |
| GDP.pdf | — | 8.0 |
| MathVision | — | 78.2 |
| MMMU-Pro | — | 72.3 |
| OmniDocBench | — | 77.9 |
| PerceptionBench | — | 36.8 |
| Video-MME v2 | — | 32.9 |
| ZeroBench Main | — | 12.0 |
| GDPval-AA v2 | — | 1211 |
| JobBench | — | 36.5 |
| τ²-Bench - Telecom | — | 73.2 |
| ARC-AGI-1 | — | 22.7 |
| ARC-AGI-2 | — | 1.9 |
| BrowseComp | — | 78.0 |
| DeepSearchQA | — | 77.7 |
Time-of-day providers use peak-hour rates; fixed-price providers use standard rates. Off-peak discounts remain in the published details.
Drag the handle above the bar for cache hits, the one below it for the input : output mix · 5% steps
Blended price = effective input rate at 90% cache hits × 80% + output rate × 20%
A blended cost needs a published input and output price on both sides, and one of these two does not have them yet.
Devstral Small 2: Listed text rates; pricing tier not confirmed.
Nemotron 3.5 Super VL: Standard pricing. DeepInfra official models API (nvidia/NVIDIA-Nemotron-3.5-Super-VL) verified 2026-11-13
| Price / 1M tokens | Devstral Small 2 | Nemotron 3.5 Super VL | Ratio |
|---|---|---|---|
| Input | — | $0.15 | — |
| Output | — | $0.50 | — |
| Cached input | — | $0.02 | — |
| Cache write | — | — | — |
| Blended | — | $0.126 | — |
Estimates use the selected token mix. Cache writes, tool fees and long-context surcharges are excluded.
Reported cost per task, paired with the score from the same run. Settings and sources may differ; costs do not affect the verdict.
| Benchmark | Devstral Small 2 | Nemotron 3.5 Super VL |
|---|---|---|
| AA Intelligence Index v5.0.1 | Not reported | |
| ARC-AGI-1 | Not reported | |
| ARC-AGI-2 | Not reported | |
| LiveBench 2.0 | Not reported | $0.1577 / task Average · Overall score (equal-weight mean of 7 suite scores, 0-100): 56.0 Max (task-specific tools) |
| Terminal-Bench 4.0 | Not reported | $0.86 / task Average · Resolution rate (%): 0.0 Unspecified Reasoning (With Tools) · vals-mini-swe |
| Terminal-Bench 4.1 | Not reported |
| Attribute | Devstral Small 2 | Nemotron 3.5 Super VL |
|---|---|---|
| Availability | Retired | Available |
| Provider | MistralAI | NVIDIA |
| Model type | Coding model | Multimodal model |
| First release | December 9, 2025 | October 15, 2026 |
| Context window | 256K tokens | 1M tokens |
| Total parameters | 24B | 120B |
| Active parameters | 24B | 12B |
| Architecture | Dense | Mixture of Experts (MoE) |
| Documented capabilities | Coding | Reasoning, Multilingual, Tool use, Function calling, Computer use, OCR, Long context, Coding |
| Knowledge cutoff | — | 2025-12 |
| Input modalities | Text | Text, Image, Video |
| Output modalities | Text | Text |
| Weights | Open weights | Open weights |
| Licence | Apache 2.0 | OpenMDW License Agreement 1.1 |
| Commercial use | Commercial use allowed | Commercial use allowed |
These are reported-score comparisons, not controlled head-to-head tests.
What people found comparing these models, or using either one. Comments also appear on the models' own pages.
Used Devstral Small 2 or Nemotron 3.5 Super VL? Sign in with Google to share what worked, what did not, and which API or setup you tested.
No community comments on either model yet.
Alternatives with the most benchmark overlap for this matchup.