G9v3-3B

AI9Stars · ai9stars/g9v3-3b

ProviderAI9Stars
Familyg9
Typellm-chat
Statusactive
Released2026-07-21
Updated2026-07-22
Parameters3B
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/sbf165.98 GB~2459298.9~9837195.6 q4
NVIDIA GB200 Grace Blackwell Superchip372.0 GB16000.0 GB/sbf165.98 GB~1873.8~7495.0 q4
AMD Instinct MI355X288.0 GB8000.0 GB/sbf165.98 GB~936.9~3747.5 q4
NVIDIA B200180.0 GB7700.0 GB/sbf165.98 GB~901.7~3607.0 q4
Google TPU7x (Ironwood)192.0 GB7380.0 GB/sbf165.98 GB~864.3~3457.1 q4
AMD Instinct MI325X256.0 GB6000.0 GB/sbf165.98 GB~702.7~2810.6 q4
AMD Instinct MI300X192.0 GB5300.0 GB/sbf165.98 GB~620.7~2482.7 q4
NVIDIA H200 SXM141.0 GB4800.0 GB/sbf165.98 GB~562.1~2248.5 q4
NVIDIA H100 NVL94.0 GB3900.0 GB/sbf165.98 GB~456.7~1826.9 q4
NVIDIA H100 SXM80.0 GB3350.0 GB/sbf165.98 GB~392.3~1569.3 q4
AMD Instinct MI250X128.0 GB3200.0 GB/sbf165.98 GB~374.8~1499.0 q4
Google TPU v5p95.0 GB2765.0 GB/sbf165.98 GB~323.8~1295.2 q4
NVIDIA A100 80GB SXM80.0 GB2039.0 GB/sbf165.98 GB~238.8~955.1 q4
NVIDIA H100 PCIe80.0 GB2000.0 GB/sbf165.98 GB~234.2~936.9 q4
NVIDIA GeForce RTX 509032.0 GB1792.0 GB/sbf165.98 GB~209.9~839.4 q4
Google TPU v6e (Trillium)32.0 GB1638.0 GB/sbf165.98 GB~191.8~767.3 q4
AMD Instinct MI21064.0 GB1600.0 GB/sbf165.98 GB~187.4~749.5 q4
NVIDIA A100 40GB SXM40.0 GB1555.0 GB/sbf165.98 GB~182.1~728.4 q4
Google TPU v432.0 GB1200.0 GB/sbf165.98 GB~140.5~562.1 q4
NVIDIA GeForce RTX 409024.0 GB1008.0 GB/sbf165.98 GB~118.0~472.2 q4
AMD Radeon RX 7900 XTX24.0 GB960.0 GB/sbf165.98 GB~112.4~449.7 q4
NVIDIA GeForce RTX 508016.0 GB960.0 GB/sbf165.98 GB~112.4~449.7 q4
NVIDIA GeForce RTX 309024.0 GB936.0 GB/sbf165.98 GB~109.6~438.5 q4
NVIDIA L40S48.0 GB864.0 GB/sbf165.98 GB~101.2~404.7 q4
Apple M3 Ultra512.0 GB819.0 GB/sbf165.98 GB~95.9~383.7 q4
AMD Radeon RX 7900 XT20.0 GB800.0 GB/sbf165.98 GB~93.7~374.8 q4
Apple M2 Ultra192.0 GB800.0 GB/sbf165.98 GB~93.7~374.8 q4
Google TPU v5e16.0 GB800.0 GB/sbf165.98 GB~93.7~374.8 q4
NVIDIA GeForce RTX 4080 SUPER16.0 GB736.0 GB/sbf165.98 GB~86.2~344.8 q4
NVIDIA GeForce RTX 4070 Ti SUPER16.0 GB672.0 GB/sbf165.98 GB~78.7~314.8 q4
AMD Radeon RX 9070 XT16.0 GB640.0 GB/sbf165.98 GB~75.0~299.8 q4
Apple M4 Max128.0 GB546.0 GB/sbf165.98 GB~63.9~255.8 q4
Apple M1 Max64.0 GB400.0 GB/sbf165.98 GB~46.8~187.4 q4
Apple M2 Max96.0 GB400.0 GB/sbf165.98 GB~46.8~187.4 q4
Apple M3 Max128.0 GB400.0 GB/sbf165.98 GB~46.8~187.4 q4
NVIDIA GeForce RTX 3060 12GB12.0 GB360.0 GB/sbf165.98 GB~42.2~168.6 q4
NVIDIA L424.0 GB300.0 GB/sbf165.98 GB~35.1~140.5 q4
NVIDIA GeForce RTX 4060 Ti 16GB16.0 GB288.0 GB/sbf165.98 GB~33.7~134.9 q4
Apple M4 Pro64.0 GB273.0 GB/sbf165.98 GB~32.0~127.9 q4
Apple M432.0 GB120.0 GB/sbf165.98 GB~14.1~56.2 q4

Where it runs

Taken from the card's availability section.

PlatformId on that platform
Hugging Faceai9stars/G9v3-3B

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_lcrG9v3-3B41.33%2026-09-10 evaluatedindependent evaluatorsource
critptG9v3-3B0.0%2026-09-10 evaluatedindependent evaluatorsource
gdpval_aaG9v3-3B15.37%2026-09-10 evaluatedindependent evaluatorsource
gpqa_diamondG9v3-3B43.84%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