Skip to content
Model Pareto

Nemotron 3 Ultra

NVIDIAOpen weightshuggingface.co ↗
Compare with nearest rivalsCompare with…
Input
$0.60/M tok
Output
$2.50/M tok
Blended
$1.075/M tok · 3:1
Speed
215tok/s
Context
262Ktokens

Open weights (550B-A55B). Price is the median third-party host price.

Rankings

Self-hostingEstimated

How we estimate →

What it would cost to run these open weights yourself on rented GPUs, compared with the API price above.

Hardware
8× B200 180GB
FP8 weights
Throughput
~21,434 tok/s
many requests batched
Price
$0.29–2.61 /M tok
busy → light use
On your own machine
Mac Studio M5 Ultra 512GB (int4, ~31 tok/s single-stream)
single consumer GPU or Mac
Assumptions (5)
  • FP8 weights (550B params, 55B active per token) + 25% KV-cache headroom ≈ 688 GB
  • 8× B200 180GB at $5.98–$14.24/GPU-hour on-demand (2026-09-24)
  • ~21,434 output tok/s aggregate at batch 613 (bandwidth-bound); MoE compute scales with active params
  • Blended 3:1 input:output; prefill ~154,636 tok/s
  • Low = 75% utilization at the low GPU price; high = 20% at the high price

Benchmark scores

~ italic, dashed = no published score yet, estimated from related benchmarks.

Coding

SWE-bench Pro (Public)Estimated
~#18 of 19 (estimated)~50.9%
  1. #1Claude Opus 5.589.9%
  2. #2Claude Sonnet 5.581.3%
  3. #17Muse Glimmer51.2%
  4. ~#18Nemotron 3 UltraEstimated50.9%
  5. #18Nex-N2.5-Mini43.8%
No published score — estimated from related benchmarks
DeepSWE v1.1Estimated
~#28 of 31 (estimated)~48.0%
  1. #1DeepSeek V4.1 Flash74.2%
  2. #2Grok 4.772.6%
  3. #27Qwen3.8 Max (0902)51.0%
  4. ~#28Nemotron 3 UltraEstimated48.0%
  5. #28Qwen3.8 27B42.2%
No published score — estimated from related benchmarks
AA Coding Agent IndexEstimated
~last of 15 (estimated)below measured range (<41.1)
  1. #1Claude Opus 5.566.0
  2. #2Claude Fable 5.162.2
  3. #14GPT-6 Luna41.1
  4. ~#15Nemotron 3 UltraEstimated<41.1
No published score — estimated from related benchmarks

Writing & chat

MMMLUEstimated
~#3 of 5 (estimated)~87.1%
  1. #1Gemini 3.1 Pro92.6%
  2. #2Gemma 4 31B88.4%
  3. ~#3Nemotron 3 UltraEstimated87.1%
  4. #3Claude Haiku 4.583.0%
No published score — estimated from related benchmarks
LMArena Text, Non-English (Elo)Verified
#35 of 531405 Elo
  1. #1Claude Opus 5.51501
  2. #2Claude Fable 5.1Few votes1496
  3. #34Mistral Medium 3.51406
  4. #35Nemotron 3 Ultra1405
  5. #36Mistral Large 31397
lmarena.ai · 2026-09-24
LMArena Text (Elo)Verified
#33 of 531426 Elo
  1. #1Claude Opus 5.51509
  2. #2Claude Fable 5.11501
  3. #33Mistral Medium 3.51426
  4. #33Nemotron 3 Ultra1426
  5. #35Muse GlimmerFew votes1424
lmarena.ai · 2026-09-24

Hard reasoning

AIME (latest)Estimated
~#5 of 25 (estimated)~91.7%
  1. #1Inkling97.1%
  2. #2Muse Glimmer94.7%
  3. #4gpt-oss-120b93.4%
  4. ~#5Nemotron 3 UltraEstimated91.7%
  5. #5Xing4.0-29B-A4B90.0%
No published score — estimated from related benchmarks

Agents

OSWorld-VerifiedEstimated
~#11 of 12 (estimated)~62.8%
  1. #1Kimi K384.8%
  2. #2Qwen3.8 27B84.3%
  3. #10Muse Glimmer65.9%
  4. ~#11Nemotron 3 UltraEstimated62.8%
  5. #11Claude Haiku 4.550.7%
No published score — estimated from related benchmarks
OSWorld 2.0Estimated
~last of 5 (estimated)below measured range (<47.5%)
  1. #1GLM-5.3 Flash59.1%
  2. #2Kimi K358.3%
  3. #4GPT-5.547.5%
  4. ~#5Nemotron 3 UltraEstimated<47.5%
No published score — estimated from related benchmarks