Skip to content
Model Pareto

Apriel-v1.6-15B-Thinker

ServiceNowOpen weightsartificialanalysis.ai ↗
Input
—
Output
—
Speed
~91tok/sest.
Context
128Ktokens

Long-tail import from Artificial Analysis (2026-09-24); AA variant: Apriel-v1.6-15B-Thinker. 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
1× B200 180GB
BF16 weights
Throughput
~7,314 tok/s
many requests batched
Price
$0.12–1.09 /M tok
busy → light use
On your own machine
RTX 5090 32GB (int4, ~170 tok/s single-stream)
single consumer GPU or Mac
Assumptions (5)
  • BF16 weights (14.86B params) + 25% KV-cache headroom ≈ 37 GB
  • 1× B200 180GB at $5.98–$14.24/GPU-hour on-demand (2026-09-24)
  • ~7,314 output tok/s aggregate at batch 209 (bandwidth-bound)
  • Blended 3:1 input:output; prefill ~35,771 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

~#12 of 15 (estimated)~61.4%
  1. #1Kimi K384.8%
  2. #2Qwen3.8 27B84.3%
  3. #11Muse Glimmer65.9%
  4. ~#12Apriel-v1.6-15B-ThinkerEstimated61.4%
  5. #12Nemotron 3.5 Super VL NVFP452.3%
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. ~#3Apriel-v1.6-15B-ThinkerEstimated53.8%
  4. #3GPT-5.6 Terra50.2%
No published score — estimated from related benchmarks
~#36 of 92 (estimated)~1128 Elo
  1. #1Claude Opus 5.51866
  2. #2Muse Wonder Full1857
  3. #35Ling-3.0-flash-VL1150
  4. ~#36Apriel-v1.6-15B-ThinkerEstimated1128
  5. #36Nex-N2-Pro1091
No published score — estimated from related benchmarks
~#50 of 91 (estimated)~14.9%
  1. #1Muse Spark 1.350.5%
  2. #2GLM-5.350.3%
  3. #49Mistral Medium 3.515.1%
  4. ~#50Apriel-v1.6-15B-ThinkerEstimated14.9%
  5. #50Gemma 4 31B14.8%
No published score — estimated from related benchmarks