Compare / head-to-head

DeepSeek V3.1vsNVIDIA Nemotron 3 Ultra 550B-A55B NVFP4

NVIDIA Nemotron 3 Ultra 550B-A55B NVFP4 leads 4 of 4 shared benchmarks. NVIDIA Nemotron 3 Ultra 550B-A55B NVFP4 has the larger context window (1M tokens).

Benchmarks from cited public sources; pricing from official pages; status from official provider feeds.

Shared benchmarks
0 – 4
NVIDIA Nemotron 3 Ultra 550B-A55B NVFP4 leads
Cheaper per token
—
list price, input + output
Larger context
NVIDIA Nemotron 3 Ultra 550B-A55B NVFP4
1M tokens
Providers
2 providers
DeepSeek · NVIDIA
DeepSeek V3.1
DeepSeek · released 2025-08-21
text
Context
128K
Max out
—
Input /1M
—
Output /1M
—
Cached /1M
—
Scores
18 · 9 core
NVIDIA Nemotron 3 Ultra 550B-A55B NVFP4
NVIDIA · released 2026-06-04
text
Context
1M
Max out
—
Input /1M
—
Output /1M
—
Cached /1M
—
Scores
22 · 5 core
Quality

Benchmark matrix

BenchmarkDeepSeek V3.1NVIDIA Nemotron 3 Ultra 550B-A55B NVFP4ΔEdge
Reported by both · 4
LMArena Elo1417.1 ↗1444.7 ↗-27.6NVIDIA Nemotron 3 Ultra 550B-A55B NVFP4
BrowseComp30% ↗41.4% ↗-11.4 ptNVIDIA Nemotron 3 Ultra 550B-A55B NVFP4
SimpleBench (SimpleBench)40% ↗41.7% ↗-1.7 ptNVIDIA Nemotron 3 Ultra 550B-A55B NVFP4
WeirdML (Håvard Tveit Ihle)38.4% ↗43.5% ↗-5.1 ptNVIDIA Nemotron 3 Ultra 550B-A55B NVFP4
Only DeepSeek V3.1 reports · 14
Aider-Polyglot (thinking)76.3% ↗not reported——
AIME 2024 (thinking)93.1% ↗not reported——
AIME 2025 (thinking)88.4% ↗not reported——
GPQA-Diamond (thinking)80.1% ↗not reported——
HMMT 2025 (thinking)84.2% ↗not reported——
Humanity's Last Exam (Python + Search)29.8% ↗not reported——
Humanity's Last Exam (thinking, text-only)15.9% ↗not reported——
LiveCodeBench (2408-2505) (thinking)74.8% ↗not reported——
MMLU-Pro (thinking)84.8% ↗not reported——
MMLU-Redux (thinking)93.7% ↗not reported——
SimpleQA (search agent)93.4% ↗not reported——
SWE Verified (Agent mode)66% ↗not reported——
SWE-bench Multilingual (Agent mode)54.5% ↗not reported——
Terminal-bench (Terminus 1 framework)31.3% ↗not reported——
Only NVIDIA Nemotron 3 Ultra 550B-A55B NVFP4 reports · 16
SWE-bench Verifiednot reported69.7% ↗——
GPQA Diamondnot reported87.9% ↗——
Humanity's Last Exam (no tools)not reported26.1% ↗——
APEX-Agents (Mercor)not reported22.7% ↗——
Chess Puzzles (Epoch AI run)not reported12% ↗epoch run——
LiveBench Agentic Coding (LiveBench)not reported38.7% ↗——
LiveBench Coding (LiveBench)not reported70.7% ↗——
LiveBench Data Analysis (LiveBench)not reported54.5% ↗——
LiveBench Instruction Following (LiveBench)not reported73.4% ↗——
LiveBench Language (LiveBench)not reported70.8% ↗——
LiveBench Mathematics (LiveBench)not reported88.7% ↗——
LiveBench Reasoning (LiveBench)not reported74.7% ↗——
Mystery Game Puzzles (Epoch AI run)not reported20% ↗epoch run——
OTIS Mock AIME 2024-2025 (Epoch AI run)not reported86.7% ↗epoch run——
Terminal-Bench 2.1not reported53.9% ↗——
Toolathlon-Verified (HKUST)not reported34.3% ↗——
Scores tagged "epoch" or "matharena" are independent runs, used only where the lab has not published its own; ⚠ marks rows MathArena flags as released after the competition. Higher is better on every row. Δ is DeepSeek V3.1 minus NVIDIA Nemotron 3 Ultra 550B-A55B NVFP4 in the benchmark's own unit. "Not reported" means the lab has not published that figure; it is not a zero. ↗ opens the source.
Specs & pricing

Side by side

SpecDeepSeek V3.1NVIDIA Nemotron 3 Ultra 550B-A55B NVFP4Edge
Context window128K tokens1M tokensNVIDIA Nemotron 3 Ultra 550B-A55B NVFP4
Max output———
Input price / 1M———
Output price / 1M———
Cached input / 1M———
Input + output / 1M
Lower is cheaper. List prices; batch, tool and regional fees excluded.
———
ModalitiestexttextTie
Released2025-08-212026-06-04—
Cited benchmark scores1822—
Reliability

Provider status

All providers →
More matchups

DeepSeek V3.1 vs …

More matchups

NVIDIA Nemotron 3 Ultra 550B-A55B NVFP4 vs …

Built by Respan
Which one wins on your data?

Public benchmarks are a starting point. Run DeepSeek V3.1 and NVIDIA Nemotron 3 Ultra 550B-A55B NVFP4 on your own prompts with Respan evals, or route to either through one gateway key with automatic failover.