K-EXAONE

LG AI Research · lgai-exaone/k-exaone-236b-a23b

ProviderLG AI Research
Familyk-exaone
Typellm-chat
Statusactive
Released2025-12-26
Updated2026-08-03
Parameters237.1B
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. This is a mixture-of-experts model and no card carries active-parameter counts, so these predictions use total parameters and understate the real speed.
DeviceMemoryBandwidthBest qualityWeightstok/sFastest
NVIDIA GB200 Grace Blackwell Superchip372.0 GB16000.0 GB/sfp8237.1 GB~47.2~94.5 q4
AMD Instinct MI355X288.0 GB8000.0 GB/sq6177.82 GB~31.5~47.2 q4
NVIDIA B200180.0 GB7700.0 GB/sq4118.55 GB~45.5~45.5 q4
Google TPU7x (Ironwood)192.0 GB7380.0 GB/sq4118.55 GB~43.6~43.6 q4
AMD Instinct MI325X256.0 GB6000.0 GB/sq6177.82 GB~23.6~35.4 q4
AMD Instinct MI300X192.0 GB5300.0 GB/sq4118.55 GB~31.3~31.3 q4
Apple M3 Ultra512.0 GB819.0 GB/sfp8237.1 GB~2.4~4.8 q4
Apple M2 Ultra192.0 GB800.0 GB/sq4118.55 GB~4.7~4.7 q4

Where it runs

Taken from the card's availability section.

PlatformId on that platform
Hugging FaceLGAI-EXAONE/K-EXAONE-236B-A23B

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
aa_lcrK-EXAONE61.33%2026-09-10 evaluatedindependent evaluatorsource
critptK-EXAONE1.12%2026-09-10 evaluatedindependent evaluatorsource
gdpval_aaK-EXAONE1.99%2026-09-10 evaluatedindependent evaluatorsource
gpqa_diamondK-EXAONE78.28%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