LG AI Research · lgai-exaone/exaone-4-5-33b
| Provider | LG AI Research |
|---|---|
| Family | exaone |
| Type | vlm |
| Status | active |
| Released | 2026-04-04 |
| Updated | 2026-07-06 |
| Parameters | 34.4B |
| Open weights | yes |
Ordered by predicted speed. Bandwidth sets decode rate; memory decides whether it runs at all.
| Device | Memory | Bandwidth | Best quality | Weights | tok/s | Fastest |
|---|---|---|---|---|---|---|
| Cerebras WSE-3 | 44.0 GB | 21000000.0 GB/s | q6 | 25.76 GB | ~570595.2 | ~855892.8 q4 |
| NVIDIA GB200 Grace Blackwell Superchip | 372.0 GB | 16000.0 GB/s | bf16 | 68.7 GB | ~163.0 | ~652.1 q4 |
| AMD Instinct MI355X | 288.0 GB | 8000.0 GB/s | bf16 | 68.7 GB | ~81.5 | ~326.1 q4 |
| NVIDIA B200 | 180.0 GB | 7700.0 GB/s | bf16 | 68.7 GB | ~78.5 | ~313.8 q4 |
| Google TPU7x (Ironwood) | 192.0 GB | 7380.0 GB/s | bf16 | 68.7 GB | ~75.2 | ~300.8 q4 |
| AMD Instinct MI325X | 256.0 GB | 6000.0 GB/s | bf16 | 68.7 GB | ~61.1 | ~244.5 q4 |
| AMD Instinct MI300X | 192.0 GB | 5300.0 GB/s | bf16 | 68.7 GB | ~54.0 | ~216.0 q4 |
| NVIDIA H200 SXM | 141.0 GB | 4800.0 GB/s | bf16 | 68.7 GB | ~48.9 | ~195.6 q4 |
| NVIDIA H100 NVL | 94.0 GB | 3900.0 GB/s | bf16 | 68.7 GB | ~39.7 | ~159.0 q4 |
| NVIDIA H100 SXM | 80.0 GB | 3350.0 GB/s | fp8 | 34.35 GB | ~68.3 | ~136.5 q4 |
| AMD Instinct MI250X | 128.0 GB | 3200.0 GB/s | bf16 | 68.7 GB | ~32.6 | ~130.4 q4 |
| Google TPU v5p | 95.0 GB | 2765.0 GB/s | bf16 | 68.7 GB | ~28.2 | ~112.7 q4 |
| NVIDIA A100 80GB SXM | 80.0 GB | 2039.0 GB/s | fp8 | 34.35 GB | ~41.6 | ~83.1 q4 |
| NVIDIA H100 PCIe | 80.0 GB | 2000.0 GB/s | fp8 | 34.35 GB | ~40.8 | ~81.5 q4 |
| NVIDIA GeForce RTX 5090 | 32.0 GB | 1792.0 GB/s | q5 | 21.47 GB | ~58.4 | ~73.0 q4 |
| Google TPU v6e (Trillium) | 32.0 GB | 1638.0 GB/s | q5 | 21.47 GB | ~53.4 | ~66.8 q4 |
| AMD Instinct MI210 | 64.0 GB | 1600.0 GB/s | fp8 | 34.35 GB | ~32.6 | ~65.2 q4 |
| NVIDIA A100 40GB SXM | 40.0 GB | 1555.0 GB/s | q6 | 25.76 GB | ~42.3 | ~63.4 q4 |
| Google TPU v4 | 32.0 GB | 1200.0 GB/s | q5 | 21.47 GB | ~39.1 | ~48.9 q4 |
| NVIDIA GeForce RTX 4090 | 24.0 GB | 1008.0 GB/s | q4 | 17.18 GB | ~41.1 | ~41.1 q4 |
| AMD Radeon RX 7900 XTX | 24.0 GB | 960.0 GB/s | q4 | 17.18 GB | ~39.1 | ~39.1 q4 |
| NVIDIA GeForce RTX 3090 | 24.0 GB | 936.0 GB/s | q4 | 17.18 GB | ~38.1 | ~38.1 q4 |
| NVIDIA L40S | 48.0 GB | 864.0 GB/s | fp8 | 34.35 GB | ~17.6 | ~35.2 q4 |
| Apple M3 Ultra | 512.0 GB | 819.0 GB/s | bf16 | 68.7 GB | ~8.3 | ~33.4 q4 |
| Apple M2 Ultra | 192.0 GB | 800.0 GB/s | bf16 | 68.7 GB | ~8.2 | ~32.6 q4 |
| Apple M4 Max | 128.0 GB | 546.0 GB/s | bf16 | 68.7 GB | ~5.6 | ~22.3 q4 |
| Apple M1 Max | 64.0 GB | 400.0 GB/s | fp8 | 34.35 GB | ~8.2 | ~16.3 q4 |
| Apple M2 Max | 96.0 GB | 400.0 GB/s | bf16 | 68.7 GB | ~4.1 | ~16.3 q4 |
| Apple M3 Max | 128.0 GB | 400.0 GB/s | bf16 | 68.7 GB | ~4.1 | ~16.3 q4 |
| NVIDIA L4 | 24.0 GB | 300.0 GB/s | q4 | 17.18 GB | ~12.2 | ~12.2 q4 |
| Apple M4 Pro | 64.0 GB | 273.0 GB/s | fp8 | 34.35 GB | ~5.6 | ~11.1 q4 |
| Apple M4 | 32.0 GB | 120.0 GB/s | q5 | 21.47 GB | ~3.9 | ~4.9 q4 |
Taken from the card's availability section.
| Platform | Id on that platform |
|---|---|
| Hugging Face | LGAI-EXAONE/EXAONE-4.5-33B |
| Benchmark | Evaluated as | Score | Evidence date | Source kind | |
|---|---|---|---|---|---|
| aa_lcr | EXAONE 4.5 33B | 54.67% | 2026-09-10 evaluated | independent evaluator | source |
| critpt | EXAONE 4.5 33B | 0.29% | 2026-09-10 evaluated | independent evaluator | source |
| gdpval_aa | EXAONE 4.5 33B | 6.38% | 2026-09-10 evaluated | independent evaluator | source |
| gpqa_diamond | EXAONE 4.5 33B | 79.39% | 2026-09-10 evaluated | independent evaluator | source |
This card reports no benchmark scores yet.
| Benchmark | Catalogue standing | Score |
|---|