Apriel-v1.5-15B-Thinker

ServiceNow · servicenow/apriel-1-5-15b-thinker

ProviderServiceNow
Familyapriel
Typevlm
Statusactive
Released2025-09-24
Updated2025-10-06
Parameters14.9B
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/sbf1629.73 GB~494488.0~1977951.9 q4
NVIDIA GB200 Grace Blackwell Superchip372.0 GB16000.0 GB/sbf1629.73 GB~376.8~1507.0 q4
AMD Instinct MI355X288.0 GB8000.0 GB/sbf1629.73 GB~188.4~753.5 q4
NVIDIA B200180.0 GB7700.0 GB/sbf1629.73 GB~181.3~725.2 q4
Google TPU7x (Ironwood)192.0 GB7380.0 GB/sbf1629.73 GB~173.8~695.1 q4
AMD Instinct MI325X256.0 GB6000.0 GB/sbf1629.73 GB~141.3~565.1 q4
AMD Instinct MI300X192.0 GB5300.0 GB/sbf1629.73 GB~124.8~499.2 q4
NVIDIA H200 SXM141.0 GB4800.0 GB/sbf1629.73 GB~113.0~452.1 q4
NVIDIA H100 NVL94.0 GB3900.0 GB/sbf1629.73 GB~91.8~367.3 q4
NVIDIA H100 SXM80.0 GB3350.0 GB/sbf1629.73 GB~78.9~315.5 q4
AMD Instinct MI250X128.0 GB3200.0 GB/sbf1629.73 GB~75.4~301.4 q4
Google TPU v5p95.0 GB2765.0 GB/sbf1629.73 GB~65.1~260.4 q4
NVIDIA A100 80GB SXM80.0 GB2039.0 GB/sbf1629.73 GB~48.0~192.0 q4
NVIDIA H100 PCIe80.0 GB2000.0 GB/sbf1629.73 GB~47.1~188.4 q4
NVIDIA GeForce RTX 509032.0 GB1792.0 GB/sfp814.86 GB~84.4~168.8 q4
Google TPU v6e (Trillium)32.0 GB1638.0 GB/sfp814.86 GB~77.1~154.3 q4
AMD Instinct MI21064.0 GB1600.0 GB/sbf1629.73 GB~37.7~150.7 q4
NVIDIA A100 40GB SXM40.0 GB1555.0 GB/sbf1629.73 GB~36.6~146.5 q4
Google TPU v432.0 GB1200.0 GB/sfp814.86 GB~56.5~113.0 q4
NVIDIA GeForce RTX 409024.0 GB1008.0 GB/sfp814.86 GB~47.5~94.9 q4
AMD Radeon RX 7900 XTX24.0 GB960.0 GB/sfp814.86 GB~45.2~90.4 q4
NVIDIA GeForce RTX 508016.0 GB960.0 GB/sq611.15 GB~60.3~90.4 q4
NVIDIA GeForce RTX 309024.0 GB936.0 GB/sfp814.86 GB~44.1~88.2 q4
NVIDIA L40S48.0 GB864.0 GB/sbf1629.73 GB~20.3~81.4 q4
Apple M3 Ultra512.0 GB819.0 GB/sbf1629.73 GB~19.3~77.1 q4
AMD Radeon RX 7900 XT20.0 GB800.0 GB/sfp814.86 GB~37.7~75.4 q4
Apple M2 Ultra192.0 GB800.0 GB/sbf1629.73 GB~18.8~75.4 q4
Google TPU v5e16.0 GB800.0 GB/sq611.15 GB~50.2~75.4 q4
NVIDIA GeForce RTX 4080 SUPER16.0 GB736.0 GB/sq611.15 GB~46.2~69.3 q4
NVIDIA GeForce RTX 4070 Ti SUPER16.0 GB672.0 GB/sq611.15 GB~42.2~63.3 q4
AMD Radeon RX 9070 XT16.0 GB640.0 GB/sq611.15 GB~40.2~60.3 q4
Apple M4 Max128.0 GB546.0 GB/sbf1629.73 GB~12.9~51.4 q4
Apple M1 Max64.0 GB400.0 GB/sbf1629.73 GB~9.4~37.7 q4
Apple M2 Max96.0 GB400.0 GB/sbf1629.73 GB~9.4~37.7 q4
Apple M3 Max128.0 GB400.0 GB/sbf1629.73 GB~9.4~37.7 q4
NVIDIA GeForce RTX 3060 12GB12.0 GB360.0 GB/sq47.43 GB~33.9~33.9 q4
NVIDIA L424.0 GB300.0 GB/sfp814.86 GB~14.1~28.3 q4
NVIDIA GeForce RTX 4060 Ti 16GB16.0 GB288.0 GB/sq611.15 GB~18.1~27.1 q4
Apple M4 Pro64.0 GB273.0 GB/sbf1629.73 GB~6.4~25.7 q4
Apple M432.0 GB120.0 GB/sfp814.86 GB~5.7~11.3 q4

Where it runs

Taken from the card's availability section.

PlatformId on that platform
Hugging FaceServiceNow-AI/Apriel-1.5-15b-Thinker

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_lcrApriel-v1.5-15B-Thinker21.33%2026-09-10 evaluatedindependent evaluatorsource
critptApriel-v1.5-15B-Thinker1.14%2026-09-10 evaluatedindependent evaluatorsource
gpqa_diamondApriel-v1.5-15B-Thinker71.31%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