DeepSeek-V4-Pro vs Nemotron 3 Ultra 550B A55B
Wins 5 of 7 areas
Coding · Agents · Reasoning · Facts · Math
Wins 2 of 7 areas
Long documents · Following instructions
DeepSeek-V4-Pro is the stronger all-rounder.Nemotron 3 Ultra 550B A55B is better at following instructions.
Scores updated · 58 tests both models report · How we compare
Where each one wins
Tests won in each of the seven areas where both have results. Each piece is one test, so a longer bar means more evidence; grey means the two scored within a point of each other.
- CodingWriting and fixing software80DeepSeek-V4-Pro8 of 9 tests · 1 tie
- AgentsCarrying out multi-step tasks on its own50DeepSeek-V4-Pro5 of 5 tests
- ReasoningHard problems that need careful thinking50DeepSeek-V4-Pro5 of 5 tests
- MathCompetition and research-level math21DeepSeek-V4-Pro2 of 4 tests · 1 tie
- FactsGetting facts right instead of making them up21DeepSeek-V4-Pro2 of 3 tests
- Following instructionsDoing exactly what it is asked02Nemotron 3 Ultra 550B A55B2 of 2 tests
- Long documentsFinding answers in very long texts01Nemotron 3 Ultra 550B A55B1 of 1 test
Long documents rests on a single test.
What it costs
Prices per million tokens, roughly 750,000 words. The bars show the cost of a million tokens read plus a million written.
DeepSeek-V4-Pro costs 60% less for the same work.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where DeepSeek-V4-Pro pulls ahead
- Finds hard-to-locate facts by browsing the webBrowseComp+39points ahead
- Multi-step tasks using many real software toolsMCP Atlas+28.9points ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+20.4points ahead
Where Nemotron 3 Ultra 550B A55B pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+64.4points ahead
- Follows detailed instructions exactlyLiveBench · Instruction Following+11.1points ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+4.9points ahead
Every test, side by side
All 58 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingDeepSeek-V4-Pro
- SciCodeDeepSeek-V4-Pro by 10.550.840.3+10.5
- Terminal-Bench 2.1DeepSeek-V4-Pro by 10.16453.9+10.1
- SWE-bench VerifiedDeepSeek-V4-Pro by 9.980.670.7+9.9
- Terminal-Bench HardDeepSeek-V4-Pro by 9.846.236.4+9.8
- SWE-bench MultilingualDeepSeek-V4-Pro by 8.576.267.7+8.5
- SWE-bench ProDeepSeek-V4-Pro by 5.755.449.7+5.7
- LiveBench · Agentic CodingDeepSeek-V4-Pro by 3.942.638.7+3.9
- LiveCodeBench v6DeepSeek-V4-Pro by 3.592.589+3.5
- LiveBench · Codingtie7070.7tie
AgentsDeepSeek-V4-Pro
- BrowseCompDeepSeek-V4-Pro by 3983.444.4+39
- MCP AtlasDeepSeek-V4-Pro by 28.973.644.7+28.9
- τ-Bench V3 · BankingDeepSeek-V4-Pro by 15.930.114.2+15.9
- Terminal-Bench 4.0DeepSeek-V4-Pro by 14.114.60.5+14.1
- GDPValDeepSeek-V4-Pro by 7.132.825.7+7.1
ReasoningDeepSeek-V4-Pro
- CritPtDeepSeek-V4-Pro by 9.812.93.1+9.8
- SimpleBenchDeepSeek-V4-Pro by 9.250.941.7+9.2
- Humanity's Last ExamDeepSeek-V4-Pro by 9.137.528.4+9.1
- LiveBench · ReasoningDeepSeek-V4-Pro by 882.774.7+8
- GPQA DiamondDeepSeek-V4-Pro by 2.188.886.7+2.1
FactsDeepSeek-V4-Pro
- AA-Omniscience · Non-hallucinationNemotron 3 Ultra 550B A55B by 64.45.970.3+64.4
- AA-Omniscience · AccuracyDeepSeek-V4-Pro by 20.44322.6+20.4
- SimpleQA VerifiedDeepSeek-V4-Pro by 13.846.232.4+13.8
MathDeepSeek-V4-Pro
- HMMT Feb. 2026DeepSeek-V4-Pro by 16.495.278.8+16.4
- IMOAnswerBenchNemotron 3 Ultra 550B A55B by 2.589.892.3+2.5
- LiveBench · MathematicsDeepSeek-V4-Pro by 290.788.7+2
- AIME 2026tie94.694.2tie
Long documentsNemotron 3 Ultra 550B A55B
- AA-LCRNemotron 3 Ultra 550B A55B by 4.674.779.3+4.6
Following instructionsNemotron 3 Ultra 550B A55B
- LiveBench · Instruction FollowingNemotron 3 Ultra 550B A55B by 11.162.473.5+11.1
- IFBenchNemotron 3 Ultra 550B A55B by 4.976.581.4+4.9
Other results29 tests, not counted
Tests outside the eight areas. They are not counted above: several are summary scores built from other tests, or the same test under another name.
- Apex (Pass@1)Nemotron 3 Ultra 550B A55B by 57.427.484.8+57.4
- Toolathlon VerifiedDeepSeek-V4-Pro by 21.655.934.3+21.6
- livebench_data_analysisDeepSeek-V4-Pro by 2074.554.5+20
- GDPval-AA v2DeepSeek-V4-Pro by 142 rating points13061164+142 rating
- τ²-Bench Telecom (AA run)DeepSeek-V4-Pro by 12.996.283.3+12.9
- τ³-Bench BankingDeepSeek-V4-Pro by 12.22613.8+12.2
- Apex-Shortlist (no tools)DeepSeek-V4-Pro by 10.985.874.9+10.9
- HLE (with tools)DeepSeek-V4-Pro by 10.848.237.4+10.8
- CritPt (no tools)DeepSeek-V4-Pro by 10.6143.4+10.6
- AA-OmniscienceNemotron 3 Ultra 550B A55B by 10.3-10.7-0.4+10.3
- Artificial Analysis Coding IndexDeepSeek-V4-Pro by 10.159.449.3+10.1
- SWEBench Pro PublicDeepSeek-V4-Pro by 955.446.4+9
- Vals.ai Financial Agent 1.1 - with web searchDeepSeek-V4-Pro by 8.662.353.7+8.6
- AA IntelligenceDeepSeek-V4-Pro by 7.530.422.9+7.5
- livebench_languageDeepSeek-V4-Pro by 7.378.170.8+7.3
- SciCode (subtask)DeepSeek-V4-Pro by 750.543.5+7
- IMOAnswerBench (with tools)Nemotron 3 Ultra 550B A55B by 6.985.492.3+6.9
- AA Agentic IndexDeepSeek-V4-Pro by 627.721.7+6
- Global-MMLU-LiteDeepSeek-V4-Pro by 3.789.385.6+3.7
- ProfBench (Search)DeepSeek-V4-Pro by 3.559.956.4+3.5
- TauBench V3 - AverageDeepSeek-V4-Pro by 2.973.270.3+2.9
- Apex-Shortlist (with tools)DeepSeek-V4-Pro by 1.786.584.8+1.7
- IOI 2025DeepSeek-V4-Pro by 15.4 rating points580.1564.7+15.4 rating
- PinchBenchNemotron 3 Ultra 550B A55B by 1.488.690+1.4
- Vals.ai Financial Agent 1.1 - without web searchNemotron 3 Ultra 550B A55B by 1.258.960.1+1.2
- TauBench V3 - Airlinetie80.880tie
- MMLU-Protie87.586.8tie
- TauBench V3 - Retailtie88.988.4tie
- StrongREJECTtie98.698.7tie
Questions people ask
Which is better, DeepSeek-V4-Pro or Nemotron 3 Ultra 550B A55B?
DeepSeek-V4-Pro wins five of the seven areas where both have results: coding, agents, reasoning, facts and math. Nemotron 3 Ultra 550B A55B wins long documents and following instructions.
Which is better for coding?
DeepSeek-V4-Pro. It wins 8 of the 9 coding tests both models report; Nemotron 3 Ultra 550B A55B wins none, and 1 is a tie.
Which is cheaper?
DeepSeek-V4-Pro costs $0.43 per million input tokens and $0.87 per million output tokens; Nemotron 3 Ultra 550B A55B costs $0.60 and $2.70. That makes DeepSeek-V4-Pro about 60% cheaper for the same work.
How do you compare the two?
We use the 58 benchmark tests both models have published scores on. The verdict counts the 29 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 29 are listed but not counted, because several are summary scores or repeat a test. Each score is the one shown on the model's own page.