Skip to content
Model Pareto

Hermes 4 - Llama-3.1 405B

Nous ResearchOpen weightsartificialanalysis.ai ↗
Input
$1.00/M tok
Output
$3.00/M tok
Blended
$1.50/M tok · 3:1
Speed
42tok/s
Context
128Ktokens

Long-tail import from Artificial Analysis (2026-09-24); AA variant: Hermes 4 - Llama-3.1 405B (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, compared with the API price above.

Hardware
4× B200 180GB
FP8 weights
Throughput
~2,555 tok/s
many requests batched
Price
$1.50–13.4 /M tok
busy → light use
On your own machine
Mac Studio M5 Ultra 512GB (int4, ~4 tok/s single-stream)
single consumer GPU or Mac
Assumptions (5)
  • FP8 weights (405.9B params) + 25% KV-cache headroom ≈ 507 GB
  • 4× B200 180GB at $5.98–$14.24/GPU-hour on-demand (2026-09-24)
  • ~2,555 output tok/s aggregate at batch 73 (bandwidth-bound)
  • Blended 3:1 input:output; prefill ~10,477 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

Hard reasoning

Agents

OSWorld 2.0Estimated
~#3 of 5 (estimated)~53.8%
  1. #1GLM-5.3 Flash59.1%
  2. #2Kimi K358.3%
  3. ~#3Hermes 4 - Llama-3.1 405BEstimated53.8%
  4. #3GPT-5.6 Terra50.2%
No published score — estimated from related benchmarks
BrowseCompEstimated
~#21 of 22 (estimated)~38.7%
  1. #1Atria Dawn Preview92.5%
  2. #2GPT-6 Astra91.5%
  3. #20Nemotron 3 Ultra44.4%
  4. ~#21Hermes 4 - Llama-3.1 405BEstimated38.7%
  5. #21Mistral Small 421.3%
No published score — estimated from related benchmarks
~#57 of 92 (estimated)~739 Elo
  1. #1Claude Opus 5.51866
  2. #2Muse Wonder Full1857
  3. #56Mistral Medium 3.5747
  4. ~#57Hermes 4 - Llama-3.1 405BEstimated739
  5. #57Ring-2.6-1T734
No published score — estimated from related benchmarks