Yi 1.5 34B
by 01.AI · yi-1.5 family
34B
parameters
text-generation code-generation reasoning multilingual math creative-writing
Yi 1.5 34B is the flagship of 01.AI's Yi 1.5 series, delivering top-tier performance in reasoning, coding, math, and multilingual tasks. With an MMLU score of 76, it competes with much larger models while remaining feasible for local deployment on high-end consumer hardware. At Q4 quantization it requires around 21 GB of VRAM, fitting on a 24 GB GPU like the RTX 4090 or a 32 GB Apple Silicon Mac. For users with the hardware to support it, the 34B offers a significant quality leap over its smaller siblings.
Quick Start with Ollama
ollama run 34b-q4_K_M | Creator | 01.AI |
| Parameters | 34B |
| Architecture | transformer-decoder |
| Context | 4K tokens |
| Released | May 13, 2024 |
| License | Apache 2.0 |
| Ollama | yi:34b |
Quantization Options
| Format | File Size | VRAM Required | Quality | Ollama Tag |
|---|---|---|---|---|
| Q4_K_M rec | 19 GB | 21 GB | | 34b-q4_K_M |
| Q8_0 | 34 GB | 36 GB | | 34b-q8_0 |
| F16 | 68 GB | 70 GB | | 34b-fp16 |
Compatible Hardware
Q4_K_M requires 21 GB VRAM
Benchmark Scores
76.0
mmlu