Apertus 8B Instruct

Swiss AI · swiss-ai/apertus-8b-instruct-2509

ProviderSwiss AI
Familyapertus
Typellm-chat
Statusactive
Released2025-08-13
Updated2026-07-17
Parameters8.1B
Open weightsyes

Lineage

RelationshipModel
Is a finetune ofswiss-ai/Apertus-8B-2509

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/sbf1616.11 GB~912665.0~3650660.1 q4
NVIDIA GB200 Grace Blackwell Superchip372.0 GB16000.0 GB/sbf1616.11 GB~695.4~2781.5 q4
AMD Instinct MI355X288.0 GB8000.0 GB/sbf1616.11 GB~347.7~1390.7 q4
NVIDIA B200180.0 GB7700.0 GB/sbf1616.11 GB~334.6~1338.6 q4
Google TPU7x (Ironwood)192.0 GB7380.0 GB/sbf1616.11 GB~320.7~1282.9 q4
AMD Instinct MI325X256.0 GB6000.0 GB/sbf1616.11 GB~260.8~1043.0 q4
AMD Instinct MI300X192.0 GB5300.0 GB/sbf1616.11 GB~230.3~921.4 q4
NVIDIA H200 SXM141.0 GB4800.0 GB/sbf1616.11 GB~208.6~834.4 q4
NVIDIA H100 NVL94.0 GB3900.0 GB/sbf1616.11 GB~169.5~678.0 q4
NVIDIA H100 SXM80.0 GB3350.0 GB/sbf1616.11 GB~145.6~582.4 q4
AMD Instinct MI250X128.0 GB3200.0 GB/sbf1616.11 GB~139.1~556.3 q4
Google TPU v5p95.0 GB2765.0 GB/sbf1616.11 GB~120.2~480.7 q4
NVIDIA A100 80GB SXM80.0 GB2039.0 GB/sbf1616.11 GB~88.6~354.5 q4
NVIDIA H100 PCIe80.0 GB2000.0 GB/sbf1616.11 GB~86.9~347.7 q4
NVIDIA GeForce RTX 509032.0 GB1792.0 GB/sbf1616.11 GB~77.9~311.5 q4
Google TPU v6e (Trillium)32.0 GB1638.0 GB/sbf1616.11 GB~71.2~284.8 q4
AMD Instinct MI21064.0 GB1600.0 GB/sbf1616.11 GB~69.5~278.1 q4
NVIDIA A100 40GB SXM40.0 GB1555.0 GB/sbf1616.11 GB~67.6~270.3 q4
Google TPU v432.0 GB1200.0 GB/sbf1616.11 GB~52.2~208.6 q4
NVIDIA GeForce RTX 409024.0 GB1008.0 GB/sbf1616.11 GB~43.8~175.2 q4
AMD Radeon RX 7900 XTX24.0 GB960.0 GB/sbf1616.11 GB~41.7~166.9 q4
NVIDIA GeForce RTX 508016.0 GB960.0 GB/sfp88.05 GB~83.4~166.9 q4
NVIDIA GeForce RTX 309024.0 GB936.0 GB/sbf1616.11 GB~40.7~162.7 q4
NVIDIA L40S48.0 GB864.0 GB/sbf1616.11 GB~37.5~150.2 q4
Apple M3 Ultra512.0 GB819.0 GB/sbf1616.11 GB~35.6~142.4 q4
AMD Radeon RX 7900 XT20.0 GB800.0 GB/sfp88.05 GB~69.5~139.1 q4
Apple M2 Ultra192.0 GB800.0 GB/sbf1616.11 GB~34.8~139.1 q4
Google TPU v5e16.0 GB800.0 GB/sfp88.05 GB~69.5~139.1 q4
NVIDIA GeForce RTX 4080 SUPER16.0 GB736.0 GB/sfp88.05 GB~64.0~127.9 q4
NVIDIA GeForce RTX 4070 Ti SUPER16.0 GB672.0 GB/sfp88.05 GB~58.4~116.8 q4
AMD Radeon RX 9070 XT16.0 GB640.0 GB/sfp88.05 GB~55.6~111.3 q4
Apple M4 Max128.0 GB546.0 GB/sbf1616.11 GB~23.7~94.9 q4
Apple M1 Max64.0 GB400.0 GB/sbf1616.11 GB~17.4~69.5 q4
Apple M2 Max96.0 GB400.0 GB/sbf1616.11 GB~17.4~69.5 q4
Apple M3 Max128.0 GB400.0 GB/sbf1616.11 GB~17.4~69.5 q4
NVIDIA GeForce RTX 3060 12GB12.0 GB360.0 GB/sfp88.05 GB~31.3~62.6 q4
NVIDIA L424.0 GB300.0 GB/sbf1616.11 GB~13.0~52.2 q4
NVIDIA GeForce RTX 4060 Ti 16GB16.0 GB288.0 GB/sfp88.05 GB~25.0~50.1 q4
Apple M4 Pro64.0 GB273.0 GB/sbf1616.11 GB~11.9~47.5 q4
Apple M432.0 GB120.0 GB/sbf1616.11 GB~5.2~20.9 q4

Where it runs

Taken from the card's availability section.

PlatformId on that platform
Hugging Faceswiss-ai/Apertus-8B-Instruct-2509

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_lcrApertus 8B Instruct0.0%2026-09-10 evaluatedindependent evaluatorsource
critptApertus 8B Instruct0.0%2026-09-10 evaluatedindependent evaluatorsource
gpqa_diamondApertus 8B Instruct25.56%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