Leaderboard / Swiss AI Initiative (Apertus)
Apertus 1.5 8B
Swiss AI Initiative (Apertus) · Apertus · released 2026-07-24 · data checked 2026-11-01 Needs verification
Apertus 1.5 8B is a 8.9B-parameter dense model from Swiss AI Initiative (Apertus) (Europe), released 2026-07-24 under the Apache-2.0 license. It scores 20.1 on the Small LM Index, #24 of 26 small models. It needs about 7.2 GB of memory at 4-bit with an 8K-token context.
Fully open: weights, training data and recipes. Trained for 1,000+ languages by EPFL, ETH Zurich and CSCS.
Benchmarks
Each bar shows this model's score. The tick marks the best small model on that benchmark. Sources: 2 reported by Swiss AI Initiative (Apertus). How we label sources.
Not reported: estimated
What the index model expects from this model's other scores. Not a measurement.
Specifications
- Lab
- Swiss AI Initiative (Apertus) (CH · Europe)
- Architecture
- Dense transformer
- Total parameters
- 8.9B
- Active parameters
- All (dense)
- Size class
- Small
- Context window
- 262K tokens
- Input modalities
- text, image, audio
- Reasoning
- Hybrid (thinking on/off)
- License
- Apache-2.0
- Openness
- Fully open: weights, training data and code
- Memory to run
- ~7.2 GB at 4-bit, 8K context (more sizes)
- Hugging Face
- 42K downloads (30 days) · 171 likes
Compare with
- Apertus 1.5 8B vs Qwen3.5 9B (9B, index 63.9)
- Apertus 1.5 8B vs MiniCPM-SALA (9.5B, index 52.6)
- Apertus 1.5 8B vs Nemotron Nano 9B v2 (8.9B, index 52.5)
- Apertus 1.5 8B vs DeepSeek-R1-0528-Qwen3-8B (8.2B, index 53.1)
Ranked in
Sources
Something wrong here? Report it.
Run it locally
No GGUF build from a well-known publisher yet.
vllm serve swiss-ai/Apertus-v1.5-8BMemory needed: weights, context cache and runtime buffers (context cache estimated: attention layout not published).
| Quantization | 4K context | 32K context | 256K context |
|---|---|---|---|
| 16-bit F16 | 19 GB | 23 GB | 54 GB |
| 8-bit Q8_0 | 11 GB | 15 GB | 46 GB |
| 4-bit Q4_K_M | 6.6 GB | 11 GB | 42 GB |