Skip to content
Model Pareto

Llama 3.1 Nemotron Ultra 253B v1

NVIDIAOpen weightsartificialanalysis.ai ↗
Compare with nearest rivalsCompare with…
Input
—
Output
—
Speed
~110tok/sest.
Context
128Ktokens

Long-tail import from Artificial Analysis (2026-09-24); AA variant: Llama 3.1 Nemotron Ultra 253B v1 (Reasoning). Price = AA median host price.

Rankings

Self-hostingEstimated

How we estimate →

What it would cost to run these open weights yourself on rented GPUs. No API sells this model, so this is its price on the chart.

Hardware
8× H100 80GB
FP8 weights
Throughput
~4,224 tok/s
many requests batched
Price
$0.88–8.40 /M tok
busy → light use
On your own machine
Mac Studio M5 Ultra 512GB (int4, ~7 tok/s single-stream)
single consumer GPU or Mac
Assumptions (5)
  • FP8 weights (253B params) + 25% KV-cache headroom ≈ 316 GB
  • 8× H100 80GB at $2.69–$6.88/GPU-hour on-demand (2026-09-24)
  • ~4,224 output tok/s aggregate at batch 123 (bandwidth-bound)
  • Blended 3:1 input:output; prefill ~14,784 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.

Writing & chat

MMMLUEstimated
~last of 5 (estimated)below measured range (<81.3%)
  1. #1Gemini 3.1 Pro92.6%
  2. #2Gemma 4 31B88.4%
  3. #4gpt-oss-120b81.3%
  4. ~#5Llama 3.1 Nemotron Ultra 253B v1Estimated<81.3%
No published score — estimated from related benchmarks
LMArena Text, Non-English (Elo)Few votesVerified
#46 of 531306 Elo
  1. #1Claude Opus 5.51501
  2. #2Claude Fable 5.1Few votes1496
  3. #45Llama 4 Maverick1311
  4. #46Llama 3.1 Nemotron Ultra 253B v1Few votes1306
  5. #47Molmo2-8BFew votes1305
lmarena.ai · 2026-09-13
LMArena Text (Elo)Few votesVerified
#45 of 531348 Elo
  1. #1Claude Opus 5.51509
  2. #2Claude Fable 5.11501
  3. #44gpt-oss-120b1352
  4. #45Llama 3.1 Nemotron Ultra 253B v1Few votes1348
  5. #46Olmo 3.1 32B Instruct1330
lmarena.ai · 2026-09-13

Hard reasoning

Agents

OSWorld-VerifiedEstimated
~#11 of 12 (estimated)~65.6%
  1. #1Kimi K384.8%
  2. #2Qwen3.8 27B84.3%
  3. #10Muse Glimmer65.9%
  4. ~#11Llama 3.1 Nemotron Ultra 253B v1Estimated65.6%
  5. #11Claude Haiku 4.550.7%
No published score — estimated from related benchmarks
OSWorld 2.0Estimated
~#3 of 5 (estimated)~53.8%
  1. #1GLM-5.3 Flash59.1%
  2. #2Kimi K358.3%
  3. ~#3Llama 3.1 Nemotron Ultra 253B v1Estimated53.8%
  4. #3GPT-5.6 Terra50.2%
No published score — estimated from related benchmarks
τ-Bench Banking (AA)Estimated
~#68 of 89 (estimated)~7.0%
  1. #1Muse Spark 1.350.5%
  2. #2GLM-5.350.3%
  3. #67LFM2.5-2.6B7.2%
  4. ~#68Llama 3.1 Nemotron Ultra 253B v1Estimated7.0%
  5. #68gpt-oss-20b7.0%
No published score — estimated from related benchmarks