Architecture
data/models.yaml. Every label is auditable
against the model's sources.
Specs
- Architecture
- dense
- Total params
- 34B
- Active params
- 34B
- Context window
- 4K tokens
- Attention
- gqa
- Position encoding
- rope
- Pretraining tokens
- 3.0T
- Post-training
- sft
- OSI-approved
- yes
- Data released
- no
- Training code
- released
Benchmarks
Each score carries the date it was published; we never infer or interpolate missing scores.
Recommended use cases
- bilingual English-Chinese chat
- commercial use under Apache 2.0
- Llama-tooling drop-in
Available quantizations
GGUF llama.cpp's container; the common local format, k-quants from Q2 to Q8.
runs on llama.cpp, Ollama
Verified via the Hugging Face model tree ↗. Community quantizations change over time; the families shown are those with published weights at audit time.
Notable innovations
- · Apache 2.0 at 34B scale from a Chinese lab
- · 3T-token bilingual pretrain
- · Llama-architecture-compatible
Lineage
Sibling to Yi-34B base; the Yi-34B-200K variant extends context to 200K with continued pretraining.