EXAONE 4.0 32B

LG AI Research · lgai-exaone/exaone-4-0-32b

ProviderLG AI Research
Familyexaone
Typellm-chat
Statusactive
Released2025-07-11
Updated2025-08-04
Parameters32B
Open weightsyes

What it fits on

Ordered by predicted speed. Bandwidth sets decode rate; memory decides whether it runs at all.

Every figure here is computed, not measured. Fit is weights at each quantisation against device memory, with a 25% allowance for the KV cache, activations and the OS. Decode rate is the memory-bandwidth roofline at 70% efficiency. Nobody has run this model on these devices.
DeviceMemoryBandwidthBest qualityWeightstok/sFastest
Cerebras WSE-344.0 GB21000000.0 GB/sfp832.0 GB~459328.8~918657.7 q4
NVIDIA GB200 Grace Blackwell Superchip372.0 GB16000.0 GB/sbf1664.01 GB~175.0~699.9 q4
AMD Instinct MI355X288.0 GB8000.0 GB/sbf1664.01 GB~87.5~350.0 q4
NVIDIA B200180.0 GB7700.0 GB/sbf1664.01 GB~84.2~336.8 q4
Google TPU7x (Ironwood)192.0 GB7380.0 GB/sbf1664.01 GB~80.7~322.8 q4
AMD Instinct MI325X256.0 GB6000.0 GB/sbf1664.01 GB~65.6~262.5 q4
AMD Instinct MI300X192.0 GB5300.0 GB/sbf1664.01 GB~58.0~231.9 q4
NVIDIA H200 SXM141.0 GB4800.0 GB/sbf1664.01 GB~52.5~210.0 q4
NVIDIA H100 NVL94.0 GB3900.0 GB/sbf1664.01 GB~42.7~170.6 q4
NVIDIA H100 SXM80.0 GB3350.0 GB/sfp832.0 GB~73.3~146.5 q4
AMD Instinct MI250X128.0 GB3200.0 GB/sbf1664.01 GB~35.0~140.0 q4
Google TPU v5p95.0 GB2765.0 GB/sbf1664.01 GB~30.2~121.0 q4
NVIDIA A100 80GB SXM80.0 GB2039.0 GB/sfp832.0 GB~44.6~89.2 q4
NVIDIA H100 PCIe80.0 GB2000.0 GB/sfp832.0 GB~43.7~87.5 q4
NVIDIA GeForce RTX 509032.0 GB1792.0 GB/sq520.0 GB~62.7~78.4 q4
Google TPU v6e (Trillium)32.0 GB1638.0 GB/sq520.0 GB~57.3~71.7 q4
AMD Instinct MI21064.0 GB1600.0 GB/sfp832.0 GB~35.0~70.0 q4
NVIDIA A100 40GB SXM40.0 GB1555.0 GB/sq624.0 GB~45.3~68.0 q4
Google TPU v432.0 GB1200.0 GB/sq520.0 GB~42.0~52.5 q4
NVIDIA GeForce RTX 409024.0 GB1008.0 GB/sq416.0 GB~44.1~44.1 q4
AMD Radeon RX 7900 XTX24.0 GB960.0 GB/sq416.0 GB~42.0~42.0 q4
NVIDIA GeForce RTX 309024.0 GB936.0 GB/sq416.0 GB~40.9~40.9 q4
NVIDIA L40S48.0 GB864.0 GB/sfp832.0 GB~18.9~37.8 q4
Apple M3 Ultra512.0 GB819.0 GB/sbf1664.01 GB~9.0~35.8 q4
Apple M2 Ultra192.0 GB800.0 GB/sbf1664.01 GB~8.7~35.0 q4
Apple M4 Max128.0 GB546.0 GB/sbf1664.01 GB~6.0~23.9 q4
Apple M1 Max64.0 GB400.0 GB/sfp832.0 GB~8.7~17.5 q4
Apple M2 Max96.0 GB400.0 GB/sbf1664.01 GB~4.4~17.5 q4
Apple M3 Max128.0 GB400.0 GB/sbf1664.01 GB~4.4~17.5 q4
NVIDIA L424.0 GB300.0 GB/sq416.0 GB~13.1~13.1 q4
Apple M4 Pro64.0 GB273.0 GB/sfp832.0 GB~6.0~11.9 q4
Apple M432.0 GB120.0 GB/sq520.0 GB~4.2~5.2 q4

Where it runs

Taken from the card's availability section.

PlatformId on that platform
Hugging FaceLGAI-EXAONE/EXAONE-4.0-32B

Verified benchmark evidence

Each row was checked against its source by a reviewer, and carries the model identifier as actually evaluated — which is not always the same as this card's.
BenchmarkEvaluated asScoreEvidence dateSource kind
critptEXAONE 4.0 32B0.0%2026-09-10 evaluatedindependent evaluatorsource
gpqa_diamondEXAONE 4.0 32B73.94%2026-09-10 evaluatedindependent evaluatorsource

Reported benchmark scores

This card reports no benchmark scores yet.

BenchmarkCatalogue standingScore

Data

This card as JSON · See it in the graph · Edit on GitHub