Skip to content
Model Pareto

Nemotron 3.5 Bolt 74B-A9B

NVIDIAOpen weightshuggingface.co ↗
Compare with nearest rivalsCompare with…
Input
—
Output
—
Speed
~170tok/sest.
Context
256Ktokens

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
1× B200 180GB
FP8 weights
Throughput
~6,254 tok/s
many requests batched
Price
$0.10–0.92 /M tok
busy → light use
On your own machine
Mac Studio M5 Max 128GB (int4, ~96 tok/s single-stream)
single consumer GPU or Mac
Assumptions (5)
  • FP8 weights (74B params, 9B active per token) + 25% KV-cache headroom ≈ 93 GB
  • 1× B200 180GB at $5.98–$14.24/GPU-hour on-demand (2026-09-24)
  • ~6,254 output tok/s aggregate at batch 179 (bandwidth-bound); MoE compute scales with active params
  • Blended 3:1 input:output; prefill ~118,125 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

Hard reasoning

Agents

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 3.5 Bolt 74B-A9BEstimated<47.5%
No published score — estimated from related benchmarks
GDPval-AA (Elo)Estimated
~#30 of 91 (estimated)~1269 Elo
  1. #1Claude Opus 5.51866
  2. #2Muse Wonder Full1857
  3. #29GPT-5.51336
  4. ~#30Nemotron 3.5 Bolt 74B-A9BEstimated1269
  5. #30MiniMax-M31242
No published score — estimated from related benchmarks
AutomationBench-AAEstimated
~#42 of 84 (estimated)~17.7%
  1. #1Muse Wonder Full71.3%
  2. #2Claude Opus 5.569.5%
  3. #40MiMo-V2.518.4%
  4. ~#42Nemotron 3.5 Bolt 74B-A9BEstimated17.7%
  5. #42Nex-N2-Pro17.6%
No published score — estimated from related benchmarks