MiniCPM-V 4.6

OpenBMB · openbmb/minicpm-v-4-6

ProviderOpenBMB
Familyminicpm
Typevlm
Statusactive
Released2026-04-13
Updated2026-08-17
Parameters1.3B
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/sbf162.6 GB~5651985.3~22607941.1 q4
NVIDIA GB200 Grace Blackwell Superchip372.0 GB16000.0 GB/sbf162.6 GB~4306.3~17225.1 q4
AMD Instinct MI355X288.0 GB8000.0 GB/sbf162.6 GB~2153.1~8612.5 q4
NVIDIA B200180.0 GB7700.0 GB/sbf162.6 GB~2072.4~8289.6 q4
Google TPU7x (Ironwood)192.0 GB7380.0 GB/sbf162.6 GB~1986.3~7945.1 q4
AMD Instinct MI325X256.0 GB6000.0 GB/sbf162.6 GB~1614.9~6459.4 q4
AMD Instinct MI300X192.0 GB5300.0 GB/sbf162.6 GB~1426.5~5705.8 q4
NVIDIA H200 SXM141.0 GB4800.0 GB/sbf162.6 GB~1291.9~5167.5 q4
NVIDIA H100 NVL94.0 GB3900.0 GB/sbf162.6 GB~1049.7~4198.6 q4
NVIDIA H100 SXM80.0 GB3350.0 GB/sbf162.6 GB~901.6~3606.5 q4
AMD Instinct MI250X128.0 GB3200.0 GB/sbf162.6 GB~861.3~3445.0 q4
Google TPU v5p95.0 GB2765.0 GB/sbf162.6 GB~744.2~2976.7 q4
NVIDIA A100 80GB SXM80.0 GB2039.0 GB/sbf162.6 GB~548.8~2195.1 q4
NVIDIA H100 PCIe80.0 GB2000.0 GB/sbf162.6 GB~538.3~2153.1 q4
NVIDIA GeForce RTX 509032.0 GB1792.0 GB/sbf162.6 GB~482.3~1929.2 q4
Google TPU v6e (Trillium)32.0 GB1638.0 GB/sbf162.6 GB~440.9~1763.4 q4
AMD Instinct MI21064.0 GB1600.0 GB/sbf162.6 GB~430.6~1722.5 q4
NVIDIA A100 40GB SXM40.0 GB1555.0 GB/sbf162.6 GB~418.5~1674.1 q4
Google TPU v432.0 GB1200.0 GB/sbf162.6 GB~323.0~1291.9 q4
NVIDIA GeForce RTX 409024.0 GB1008.0 GB/sbf162.6 GB~271.3~1085.2 q4
AMD Radeon RX 7900 XTX24.0 GB960.0 GB/sbf162.6 GB~258.4~1033.5 q4
NVIDIA GeForce RTX 508016.0 GB960.0 GB/sbf162.6 GB~258.4~1033.5 q4
NVIDIA GeForce RTX 309024.0 GB936.0 GB/sbf162.6 GB~251.9~1007.7 q4
NVIDIA L40S48.0 GB864.0 GB/sbf162.6 GB~232.5~930.2 q4
Apple M3 Ultra512.0 GB819.0 GB/sbf162.6 GB~220.4~881.7 q4
AMD Radeon RX 7900 XT20.0 GB800.0 GB/sbf162.6 GB~215.3~861.3 q4
Apple M2 Ultra192.0 GB800.0 GB/sbf162.6 GB~215.3~861.3 q4
Google TPU v5e16.0 GB800.0 GB/sbf162.6 GB~215.3~861.3 q4
NVIDIA GeForce RTX 4080 SUPER16.0 GB736.0 GB/sbf162.6 GB~198.1~792.4 q4
NVIDIA GeForce RTX 4070 Ti SUPER16.0 GB672.0 GB/sbf162.6 GB~180.9~723.5 q4
AMD Radeon RX 9070 XT16.0 GB640.0 GB/sbf162.6 GB~172.3~689.0 q4
Apple M4 Max128.0 GB546.0 GB/sbf162.6 GB~147.0~587.8 q4
Apple M1 Max64.0 GB400.0 GB/sbf162.6 GB~107.7~430.6 q4
Apple M2 Max96.0 GB400.0 GB/sbf162.6 GB~107.7~430.6 q4
Apple M3 Max128.0 GB400.0 GB/sbf162.6 GB~107.7~430.6 q4
NVIDIA GeForce RTX 3060 12GB12.0 GB360.0 GB/sbf162.6 GB~96.9~387.6 q4
NVIDIA L424.0 GB300.0 GB/sbf162.6 GB~80.7~323.0 q4
NVIDIA GeForce RTX 4060 Ti 16GB16.0 GB288.0 GB/sbf162.6 GB~77.5~310.1 q4
Apple M4 Pro64.0 GB273.0 GB/sbf162.6 GB~73.5~293.9 q4
Apple M432.0 GB120.0 GB/sbf162.6 GB~32.3~129.2 q4

Where it runs

Taken from the card's availability section.

PlatformId on that platform
Hugging Faceopenbmb/MiniCPM-V-4.6

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_lcrMiniCPM-V 4.6 1.3B7.0%2026-09-10 evaluatedindependent evaluatorsource
critptMiniCPM-V 4.6 1.3B0.0%2026-09-10 evaluatedindependent evaluatorsource
gdpval_aaMiniCPM-V 4.6 1.3B0.0%2026-09-10 evaluatedindependent evaluatorsource
gpqa_diamondMiniCPM-V 4.6 1.3B30.51%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