Skip to content
Model Pareto

Qwen3.8 27B MXFP4

Red Hat AIOpen weightsMXFP4quantized from Qwen3.8 27Bhuggingface.co ↗
Compare with nearest rivalsCompare with…
Input
—
Output
—
Speed
~82tok/sest.
Context
262Ktokens

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
MXFP4 weights
Throughput
~5,953 tok/s
many requests batched
Price
$0.14–1.22 /M tok
busy → light use
On your own machine
RTX 5090 32GB (MXFP4, ~85 tok/s single-stream)
single consumer GPU or Mac
Assumptions (5)
  • MXFP4 weights (27.8B params) + 25% KV-cache headroom ≈ 18 GB
  • 1× B200 180GB at $5.98–$14.24/GPU-hour on-demand (2026-09-24)
  • ~5,953 output tok/s aggregate at batch 170 (bandwidth-bound)
  • Blended 3:1 input:output; prefill ~38,242 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.

Hard reasoning

Humanity's Last ExamEstimated
~#54 of 185 (estimated)~26.4%
  1. #1Claude Opus 5.561.4%
  2. #2Claude Fable 5.159.1%
  3. #53MiMo-V2.527.2%
  4. ~#54Qwen3.8 27B MXFP4Estimated26.4%
  5. #54Solar Mini 425.8%
No published score — estimated from related benchmarks
AA Intelligence IndexEstimated
~#32 of 86 (estimated)~27.2
  1. #1Claude Opus 5.557.6
  2. #2Claude Sonnet 5.556.0
  3. #31MiniMax-M329.2
  4. ~#32Qwen3.8 27B MXFP4Estimated27.2
  5. #32Quasar 438B26.7
No published score — estimated from related benchmarks