Sorted by Hugging Face repo date
Qwen3.8 27B is the newest catalog entry, added on Hugging Face 2026-08-05. On a 16 GB RTX 5070 Ti it is offload at 8K context.
| Added | Model | Quant | File | On 16 GB 5070 Ti |
|---|---|---|---|---|
| 2026-08-05 | Qwen3.8 27B | UD-Q4_K_M | 15.3 GiB | Offload |
| 2026-05-23 | Gemma 4 12B | Q4_K_M | 6.63 GiB | Fits |
| 2026-04-15 | Qwen3.6 35B-A3B | UD-Q4_K_M | 20.6 GiB | Offload |
| 2026-03-11 | Gemma 4 31B | Q4_K_M | 17.1 GiB | Offload |
| 2026-03-11 | Gemma 4 26B-A4B | UD-Q4_K_M | 15.8 GiB | Offload |
| 2026-03-02 | Gemma 4 E4B | Q4_K_M | 4.64 GiB | Fits |
| 2026-03-02 | Gemma 4 E2B | Q4_K_M | 2.89 GiB | Fits |
| 2026-02-27 | Qwen3.5 9B | Q4_K_M | 5.29 GiB | Fits |
| 2025-11-28 | Devstral Small 2 24B Instruct | Q4_K_M | 13.3 GiB | Tight |
| 2025-08-04 | gpt-oss-120b | MXFP4 | 59.0 GiB | No |
| 2025-08-04 | gpt-oss-20b | MXFP4 | 11.3 GiB | Fits |
| 2025-07-08 | SmolLM3 3B | Q4_K_M | 1.78 GiB | Fits |
| 2025-06-19 | Mistral Small 3.2 24B Instruct | Q4_K_M | 13.3 GiB | Tight |
| 2025-04-27 | Qwen3 30B-A3B | Q4_K_M | 17.3 GiB | Offload |
| 2025-04-27 | Qwen3 32B | Q4_K_M | 18.4 GiB | Offload |
| 2025-04-27 | Qwen3 14B | Q4_K_M | 8.38 GiB | Fits |
| 2025-04-27 | Qwen3 8B | Q4_K_M | 4.68 GiB | Fits |
| 2025-03-01 | Gemma 3 27B Instruct | Q4_K_M | 15.4 GiB | Offload |
| 2025-03-01 | Gemma 3 12B Instruct | Q4_K_M | 6.80 GiB | Fits |
| 2025-02-20 | Gemma 3 4B Instruct | Q4_K_M | 2.32 GiB | Fits |
| 2025-02-19 | Phi-4-mini Instruct | Q4_K_M | 2.32 GiB | Fits |
| 2025-01-20 | DeepSeek-R1 Distill Qwen 32B | Q4_K_M | 18.5 GiB | Offload |
| 2025-01-20 | DeepSeek-R1 Distill Qwen 14B | Q4_K_M | 8.37 GiB | Fits |
| 2024-12-11 | Phi-4 | Q4_K_M | 8.43 GiB | Fits |
| 2024-11-26 | Llama 3.3 70B Instruct | Q4_K_M | 39.6 GiB | No |
| 2024-09-18 | Llama 3.2 3B Instruct | Q4_K_M | 1.88 GiB | Fits |
| 2024-09-18 | Llama 3.2 1B Instruct | Q4_K_M | 0.75 GiB | Fits |
| 2024-09-17 | Qwen2.5 32B Instruct | Q4_K_M | 18.5 GiB | Offload |
| 2024-09-16 | Qwen2.5 72B Instruct | Q4_K_M | 44.2 GiB | No |
| 2024-09-16 | Qwen2.5 14B Instruct | Q4_K_M | 8.37 GiB | Fits |
| 2024-09-16 | Qwen2.5 7B Instruct | Q4_K_M | 4.36 GiB | Fits |
| 2024-07-18 | Llama 3.1 8B Instruct | Q4_K_M | 4.58 GiB | Fits |
| 2024-07-17 | Mistral Nemo 12B Instruct | Q4_K_M | 6.96 GiB | Fits |
FAQ
What does new mean?
The Hugging Face repo creation date of the official model, not a review date. GGUF repos are often created later and are not the sort key.
How fast can a model show up?
A model is added when its config.json and a single-file GGUF size can both be cited. A name without those files waits.
Why is the 5070 Ti the reference?
It is a current 16 GB gaming card, which is the common 'can I run the new one' question. The model page has the rest of the machines.