bartowski
bottlecapai ThinkingCap gemma 4 26B A4B it
bottlecapai ThinkingCap gemma 4 26B A4B it is a 25.2B mixture-of-experts gemma4 model from bartowski, with about 4.0B parameters active per token. Configured context length is 262,144 tokens. License: apache-2.0. Quant sizes are the GGUF file bytes published on Hugging Face when a file exists.
Parameters25.2B / 4.0B active
Context256K
Attentionhybrid
Licenseapache-2.0
30 layers · 8 KV heads · head dim 256 · 25 sliding layers · GGUF repo · config
Quant file sizes
| Quant | File size | Source |
|---|---|---|
| BF16 | 47.0 GB | Hugging Face file bytes |
| Q8_0 | 25.0 GB | Hugging Face file bytes |
| Q6_K | 66.5 GB | Hugging Face file bytes |
| Q5_K_M | 18.6 GB | Hugging Face file bytes |
| Q5_K_S | 17.1 GB | Hugging Face file bytes |
| Q4_K_M | 16.7 GB | Hugging Face file bytes |
| Q4_K_S | 15.0 GB | Hugging Face file bytes |
| IQ4_NL | 14.6 GB | Hugging Face file bytes |
| IQ4_XS | 13.3 GB | Hugging Face file bytes |
| Q4_1 | 15.0 GB | Hugging Face file bytes |
| Q4_0 | 13.7 GB | Hugging Face file bytes |
| Q3_K_L | 13.8 GB | Hugging Face file bytes |
| Q3_K_M | 12.8 GB | Hugging Face file bytes |
| Q3_K_S | 11.5 GB | Hugging Face file bytes |
| IQ3_XXS | 11.7 GB | Hugging Face file bytes |
| Q2_K | 10.8 GB | Hugging Face file bytes |
| IQ2_S | 9.60 GB | Hugging Face file bytes |
On common machines
| Device | Verdict | Quant | Speed |
|---|---|---|---|
| GeForce RTX 4060 Ti 8GB | Expert offload | Q4_K_M | 47–64 |
| GeForce RTX 3060 12GB | Tight fit | IQ2_S | 54–74 |
| GeForce RTX 4060 Ti 16GB | Tight fit | IQ4_XS | 37–51 |
| GeForce RTX 4090 | Tight fit | Q5_K_M | 107–146 |
| GeForce RTX 5090 | Tight fit | Q8_0 | 157–214 |
| Apple M4 Pro | Tight fit | Q4_K_M | 39–53 |
| Apple M5 | Tight fit | IQ2_S | 23–31 |
| CPU only · DDR5-5600 | Runs on CPU | Q5_K_M | 9.5–13 |
Smallest devices that fit it fully
- GeForce RTX 5070 — Tight fit, IQ2_S, 101–138 tok/s
- GeForce RTX 4070 Ti — Tight fit, IQ2_S, 76–104 tok/s
- GeForce RTX 4070 Super — Tight fit, IQ2_S, 76–104 tok/s
- GeForce RTX 4070 — Tight fit, IQ2_S, 76–104 tok/s
- GeForce RTX 3080 Ti — Tight fit, IQ2_S, 138–188 tok/s
- GeForce RTX 3080 12GB — Tight fit, IQ2_S, 138–188 tok/s
Run it
Commands for the GeForce RTX 4060 Ti 16GB estimate (IQ4_XS).
llama.cpp
llama-server -hf bartowski/bottlecapai_ThinkingCap-gemma-4-26B-A4B-it-GGUF:IQ4_XS -c 8192 -ngl 30
No matching name in the Ollama library snapshot.
LM Studio
lms get bartowski/bottlecapai_ThinkingCap-gemma-4-26B-A4B-it-GGUF # Then load quant IQ4_XS