Skip to content
Model Pareto

EXAONE 4.5 33B

LG AI Research (EXAONE)Open weightsartificialanalysis.ai ↗
Input
—
Output
—
Speed
~83tok/sest.
Context
262Ktokens

Long-tail import from Artificial Analysis (2026-09-24); AA variant: EXAONE 4.5 33B. 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
FP8 weights
Throughput
~4,788 tok/s
many requests batched
Price
$0.17–1.49 /M tok
busy → light use
On your own machine
RTX 5090 32GB (int4, ~76 tok/s single-stream)
single consumer GPU or Mac
Assumptions (5)
  • FP8 weights (33B params) + 25% KV-cache headroom ≈ 41 GB
  • 1× B200 180GB at $5.98–$14.24/GPU-hour on-demand (2026-09-24)
  • ~4,788 output tok/s aggregate at batch 137 (bandwidth-bound)
  • Blended 3:1 input:output; prefill ~32,216 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

Vision

MMMUEstimated
~#2 of 3 (estimated)~71.3%
  1. #1Claude Haiku 4.573.2%
  2. ~#2EXAONE 4.5 33BEstimated71.3%
  3. #2Llama 4 Maverick69.4%
No published score — estimated from related benchmarks
~#11 of 12 (estimated)~75.2%
  1. #1Gemini 3.8 Flash86.2%
  2. #2Kimi K384.8%
  3. #10Claude Sonnet 577.0%
  4. ~#11EXAONE 4.5 33BEstimated75.2%
  5. #11Gemini 3.5 Flash-Lite74.5%
No published score — estimated from related benchmarks
~#5 of 8 (estimated)~81.8%
  1. #1MiniMax-M391.6%
  2. #2Kimi K391.1%
  3. #4Gemini 3.1 Pro85.3%
  4. ~#5EXAONE 4.5 33BEstimated81.8%
  5. #5Claude Haiku 4.579.6%
No published score — estimated from related benchmarks
GDP.pdf (AA)Estimated
~#34 of 48 (estimated)~9.8%
  1. #1Muse Wonder Full32.6%
  2. #2GPT-6 Astra31.0%
  3. #33Muse Glimmer10.0%
  4. ~#34EXAONE 4.5 33BEstimated9.8%
  5. #34MiniMax-M39.8%
No published score — estimated from related benchmarks
ChartographyEstimated
~#22 of 25 (estimated)~10.9%
  1. #1Claude Fable 5.273.4%
  2. #2GPT-6 Astra71.0%
  3. #21DeepSeek V4.1 Flash11.6%
  4. ~#22EXAONE 4.5 33BEstimated10.9%
  5. #22Gemini 3.5 Flash-Lite10.2%
No published score — estimated from related benchmarks

Hard reasoning

Agents

~#11 of 15 (estimated)~66.4%
  1. #1Kimi K384.8%
  2. #2Qwen3.8 27B84.3%
  3. #10Nex-N2.5-Mini71.2%
  4. ~#11EXAONE 4.5 33BEstimated66.4%
  5. #11Muse Glimmer65.9%
No published score — estimated from related benchmarks
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. ~#5EXAONE 4.5 33BEstimated<47.5%
No published score — estimated from related benchmarks
BrowseCompEstimated
~#19 of 22 (estimated)~66.8%
  1. #1Atria Dawn Preview92.5%
  2. #2GPT-6 Astra91.5%
  3. #18Nemotron 3.5 Bolt 74B-A9B70.0%
  4. ~#19EXAONE 4.5 33BEstimated66.8%
  5. #19Mistral Medium 3.548.6%
No published score — estimated from related benchmarks
~#45 of 85 (estimated)~17.1%
  1. #1Claude Fable 5.276.4%
  2. #2Muse Wonder Full71.3%
  3. #44Qwen3.7 Plus17.4%
  4. ~#45EXAONE 4.5 33BEstimated17.1%
  5. #45Ling-3.0-flash-VL15.7%
No published score — estimated from related benchmarks
~#46 of 91 (estimated)~16.8%
  1. #1Muse Spark 1.350.5%
  2. #2GLM-5.350.3%
  3. #45Nemotron 3.5 Bolt 74B-A9B17.0%
  4. ~#46EXAONE 4.5 33BEstimated16.8%
  5. #46A.X-K216.1%
No published score — estimated from related benchmarks