Hy3

Tencent · tencent/hy3

ProviderTencent
Familyhy3
Typellm-chat
Statusactive
Released2026-07-02
Updated2026-07-20
Parameters298.8B
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/sq6224.09 GB~50.0~75.0 q4
AMD Instinct MI355X288.0 GB8000.0 GB/sq5186.74 GB~30.0~37.5 q4
AMD Instinct MI325X256.0 GB6000.0 GB/sq5186.74 GB~22.5~28.1 q4
Apple M3 Ultra512.0 GB819.0 GB/sfp8298.79 GB~1.9~3.8 q4

Where it runs

Taken from the card's availability section.

PlatformId on that platform
Hugging Facetencent/Hy3

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_lcrHy379.0%2026-09-10 evaluatedindependent evaluatorsource
arena_elo_style_controlhy31455.39 elo2026-09-10 evaluatedindependent evaluatorsource
critptHy34.86%2026-09-10 evaluatedindependent evaluatorsource
gdpval_aaHy331.8%2026-09-10 evaluatedindependent evaluatorsource
gpqa_diamondHy389.7%2026-09-10 evaluatedindependent evaluatorsource
scicodeHy348.61%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