Skip to content
Model Pareto

DeepSeek-V4-Pro-0813-nvfp4-DSpark

NVIDIAOpen weightsNVFP4quantized from DeepSeek V4 Pro (0813)huggingface.co ↗
Compare with nearest rivalsCompare with…
Input
—
Output
—
Speed
~130tok/sest.
Context
1Mtokens

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
16× B200 180GB
NVFP4 weights
Throughput
~43,073 tok/s
many requests batched
Price
$0.30–2.64 /M tok
busy → light use
On your own machine
Too large
needs data-center GPUs
Assumptions (5)
  • NVFP4 weights (1650B params, 49B active per token) + 40% KV-cache headroom ≈ 1299 GB
  • 16× B200 180GB at $5.98–$14.24/GPU-hour on-demand (2026-09-24)
  • ~43,073 output tok/s aggregate at batch 1450 (bandwidth-bound); MoE compute scales with active params
  • Blended 3:1 input:output; prefill ~295,071 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 VerifiedEstimated
~#3 of 14 (estimated)~84.8%
  1. #1Claude Opus 596.0%
  2. #2Claude Sonnet 585.2%
  3. ~#3DeepSeek-V4-Pro-0813-nvfp4-DSparkEstimated84.8%
  4. #3Gemini 3.1 Pro80.6%
No published score — estimated from related benchmarks
AA Coding Agent IndexEstimated
~#11 of 18 (estimated)~52.1
  1. #1Claude Opus 5.566.0
  2. #2Gemini 4 Argon63.8
  3. #10GLM-5.353.6
  4. ~#11DeepSeek-V4-Pro-0813-nvfp4-DSparkEstimated52.1
  5. #11Kimi K351.9
No published score — estimated from related benchmarks

Hard reasoning

Humanity's Last ExamEstimated
~#46 of 185 (estimated)~30.3%
  1. #1Claude Opus 5.561.4%
  2. #2Claude Fable 5.159.1%
  3. #45Inkling31.9%
  4. ~#46DeepSeek-V4-Pro-0813-nvfp4-DSparkEstimated30.3%
  5. #46A.X-K229.6%
No published score — estimated from related benchmarks
AIME (latest)Estimated
~#2 of 32 (estimated)~98.9%
  1. #1GLM-5.3 NVFP499.5%
  2. ~#2DeepSeek-V4-Pro-0813-nvfp4-DSparkEstimated98.9%
  3. #2Inkling97.1%
No published score — estimated from related benchmarks

Agents

OSWorld-VerifiedEstimated
~#5 of 12 (estimated)~82.0%
  1. #1Kimi K384.8%
  2. #2Qwen3.8 27B84.3%
  3. #4MiMo-V2.6-Pro82.0%
  4. ~#5DeepSeek-V4-Pro-0813-nvfp4-DSparkEstimated82.0%
  5. #5Claude Sonnet 581.2%
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. ~#3DeepSeek-V4-Pro-0813-nvfp4-DSparkEstimated53.8%
  4. #3GPT-5.6 Terra50.2%
No published score — estimated from related benchmarks
BrowseCompEstimated
~#12 of 18 (estimated)~84.1%
  1. #1Atria Dawn Preview92.5%
  2. #2GPT-6 Astra91.5%
  3. #11Hy384.2%
  4. ~#12DeepSeek-V4-Pro-0813-nvfp4-DSparkEstimated84.1%
  5. #12MiniMax-M383.5%
No published score — estimated from related benchmarks
GDPval-AA (Elo)Estimated
~#16 of 88 (estimated)~1563 Elo
  1. #1Claude Opus 5.51866
  2. #2Claude Fable 5.11758
  3. #15Step 5 Preview1566
  4. ~#16DeepSeek-V4-Pro-0813-nvfp4-DSparkEstimated1563
  5. #16GPT-6 Astra1542
No published score — estimated from related benchmarks