Request a pilot
‹ PublicAI Index

The LLM benchmark aggregator.

Nemotron 3 Ultra 550B A55B

NVIDIA · 550B · A55B

Strongest in Instruction following (#14 of 57), weakest in Safety (#256 of 337). Above par in 5 of 19 scopes. Among the models it meets almost everywhere, it finishes behind Grok 4.7 and GPT-6 Astra and ahead of Grok Build 0.1 and Command A.

§ 1 · Profile

What it is good at

Bars run from 50 — the average of the models each source lists — so right of the line is above par. The middle column is the gap to whoever leads that scope.

Agents52.6#91/268
Knowledge work53.4#61/178
Reasoning44#134/178
Mathematics49.3#83/140
Reasoning37.4#126/140
Coding39.1#159/165
Code generation36.3#67/77
Agentic coding40.9#138/157
Core abilities39.4#190/204
Instruction following59.6#14/57
Language34#54/57
Data analysis26#55/57
General intelligence38.1#185/204
Safety47#256/337
Secure code63.5#26/274
Toxicity avoidance52.4#167/272
Fairness43.9#222/300
Jailbreak resistance32.9#246/272
Harm refusal44.9#250/300

§ 2 · Head to head

What it beats, and what beats it

The same models turn up scope after scope. Counted once: where both were placed, who finished higher. Bars run right for Nemotron 3 Ultra 550B A55B, left for the other.

Qwen3.6 27Bwon 9 · lost 4
Grok 4.3won 13 · lost 6
Grok 4.1 Fastwon 8 · lost 4
Gemini 3.5 Flash Litewon 8 · lost 5
GPT-5.4 Miniwon 8 · lost 6
GPT-4o Miniwon 9 · lost 7
Command Awon 7 · lost 6
Grok Build 0.1won 10 · lost 9
DeepSeek V3won 8 · lost 8
GPT-4owon 5 · lost 7
Claude 3.5 Haikuwon 5 · lost 7
GPT-4.1 Nanowon 6 · lost 9
Llama 4 Scout Instructwon 5 · lost 8
Gemini 2.0 Flashwon 5 · lost 8
DeepSeek R1won 6 · lost 10
O3 Miniwon 5 · lost 9
GPT-5.4 Nanowon 5 · lost 9
GLM-4.6won 5 · lost 9
Gemma 3 27B Itwon 4 · lost 8
GPT OSS 20Bwon 4 · lost 8
GPT OSS 120Bwon 5 · lost 10
GPT-4.1 Miniwon 5 · lost 10
Gemini 3.5 Flashwon 6 · lost 13
Gemma 4 12Bwon 4 · lost 10
GPT-4.1won 4 · lost 10
Gemma 4 31Bwon 4 · lost 10
Grok 4won 4 · lost 10
DeepSeek V3.2won 4 · lost 11
Kimi K2.6won 5 · lost 14
MiniMax M3won 5 · lost 14
DeepSeek V4 Flashwon 5 · lost 14
Claude 3.5 Sonnetwon 3 · lost 9
Claude 3.7 Sonnetwon 3 · lost 9
Gemini 2.5 Prowon 4 · lost 12
DeepSeek V4 Flash Visionwon 3 · lost 10
Kimi K2won 3 · lost 11
Kimi K2.7 Codewon 4 · lost 15
GLM-5.3 Flashwon 4 · lost 15
GLM-5.2won 4 · lost 15
Gemini 3.6 Flashwon 4 · lost 15
Grok 4.5won 4 · lost 15
Gemini 2.5 Flashwon 3 · lost 12
O3won 3 · lost 12
GPT-5 Miniwon 3 · lost 12
Claude Sonnet 4won 3 · lost 13
MiniMax M2.7won 2 · lost 10
GPT-5.6 Lunawon 3 · lost 16
Muse Spark 1.1won 3 · lost 16
DeepSeek V4 Prowon 3 · lost 16
GPT-6 Solwon 3 · lost 16
GPT-5.6 Solwon 3 · lost 16
GPT-5.1won 2 · lost 11
Qwen3.8 27Bwon 2 · lost 11
GLM-5.3won 2 · lost 11
Mimo V2.5 Prowon 2 · lost 12
Claude Opus 4won 2 · lost 12
Gemini 3.1 Prowon 2 · lost 12
Claude Haiku 4.5won 2 · lost 13
Claude Sonnet 4.5won 2 · lost 13
Claude Opus 4.6won 2 · lost 13
O4 Miniwon 2 · lost 13
GPT-5won 2 · lost 13
Claude Sonnet 4.6won 2 · lost 15
GPT-6 Lunawon 2 · lost 17
Inklingwon 2 · lost 17
Claude Sonnet 5won 2 · lost 17
Qwen3.7 Maxwon 2 · lost 17
Claude Opus 4.8won 2 · lost 17
Claude Opus 4.7won 2 · lost 17
GPT-5.6 Terrawon 2 · lost 17
Gemini 3.8 Flashwon 2 · lost 17
GPT-5.5won 2 · lost 17
GLM-5.1won 1 · lost 12
Qwen3.6 Pluswon 1 · lost 12
Claude Opus 4.5won 1 · lost 12
Kimi K2 Thinkingwon 1 · lost 12
Qwen3.8 Flash Nextwon 1 · lost 12
DeepSeek V4.1 Flashwon 1 · lost 12
Kimi K3won 1 · lost 12
Claude Fable 5.1won 1 · lost 12
Qwen3 235B A22B Instructwon 1 · lost 13
Mimo V2.6 Prowon 1 · lost 13
Kimi K2.5won 1 · lost 14
GPT-5.2won 1 · lost 17
Claude Opus 5won 1 · lost 18
GPT-5.4won 1 · lost 18
Grok 4.6won 1 · lost 18
Qwen3.8 Maxwon 1 · lost 18
Claude Opus 5.5won 1 · lost 18
GPT-5 Nanowon 0 · lost 12
Muse Spark 1.2won 0 · lost 13
Gemini 3.7 Flashwon 0 · lost 13
Muse Spark 1.3won 0 · lost 13
Claude Fable 5won 0 · lost 13
GPT-6 Astrawon 0 · lost 14
Grok 4.7won 0 · lost 19

§ 3 · Sources

Where the numbers come from

4 publications, 16 figures. Every one links to the page it was read from.

GDPval-AA ↗1 measureread 2026-09-26
GDPval-AA 1000
AA-Briefcase ↗1 measureread 2026-09-26
AA-Briefcase 876
LiveBench ↗8 measuresread 2026-09-26
LiveBench 67.4LiveBench · Reasoning 74.7LiveBench · Coding 70.7LiveBench · Agentic Coding 38.7LiveBench · Mathematics 88.7LiveBench · Data Analysis 54.5LiveBench · Language 70.8LiveBench · Instruction Following 73.4
Enkrypt AI Safety Leaderboard ↗6 measuresread 2026-09-26
Enkrypt · Jailbreak risk 27.2%Enkrypt · Harmful content risk 9.4%Enkrypt · CBRN risk 52.7%Enkrypt · Toxicity risk 4.6%Enkrypt · Bias risk 87.3%Enkrypt · Insecure code risk 6.2%

Badge

PublicAI Index badge for Nemotron 3 Ultra 550B A55B[![PublicAI Index](https://publicai.io/model-index/badge?model=nemotron-3-ultra-550b-a55b)](https://publicai.io/model-index/m/nemotron-3-ultra-550b-a55b)