Performance ranking (third-party evaluations)
Performance ranking of models that run in 8 GB of VRAM (Epoch Capabilities Index (Epoch AI))
Epoch AI's capability index (Epoch Capabilities Index), which combines results from many benchmarks. The range is the 95% interval.
Only models with a quantization whose estimated memory fits in 8 GB at an 8,192-token context.
Top 7 of 7 models listed on this site
Values as of Aug 14, 2026
-
1Qwen3.8-27B
Qwen 27.3B Qwen 3.8 27B
Index (95% interval) 152.7 150.7–154.9
UD-IQ1_M 8.47 GB
-
2Qwen3.5-9B
Qwen 9B Qwen3.5-9B
Index (95% interval) 139.4 137.1–141.5
Q5_K_M 8.48 GB
-
3Qwen3-14B
Qwen 14.8B Qwen3-14B
Index (95% interval) 136.2 133.5–138.7
IQ3_XS 8.57 GB
-
4Qwen3-8B
Qwen 8.2B Qwen3-8B
Index (95% interval) 132.2 129.1–135.1
Q5_K_L 8.29 GB
-
5phi-4
microsoft 14.7B Phi-4
Index (95% interval) 129.1 126.1–131.9
Q2_K 7.97 GB
-
6Qwen2.5-Coder-7B
QuantFactory 7.6B Qwen2.5-Coder (7B)
Index (95% interval) 108.0 98.7–116.4
Q6_K 7.57 GB
-
7Mistral-7B-v0.1
mistralai 7.2B Mistral 7B v0.1
Index (95% interval) 103.3 95.7–109.8
Q6_K 7.62 GB
Estimated Memory at 8K: Estimated with a formula. Actual usage depends on your setup and settings.
About this ranking
- Values for open-weight models come from Epoch AI's data (Epoch Capabilities Index). Updated weekly.
- Values are for the original model before quantization. GGUF quantization (especially small quantizations) can lower performance.
- Values are attached only when the model name in the source matches a model listed on this site (its original Hugging Face repository). Fine-tunes and modified versions do not get them.
- Each source is a separate ranking. We do not mix values from different sources or assign our own scores.
- Small and new models are often not evaluated by the source, so they do not appear here.
Values are from Epoch Capabilities Index (Epoch AI) data (CC BY 4.0), shown unchanged (sorting and filtering by this site).