Leaderboard / Kakao (Kanana)
Kanana-2 30B-A3B Thinking
Kakao (Kanana) · Kanana-2 · released 2026-01-14 · data checked 2026-11-01
Kanana-2 30B-A3B Thinking is a 30.7B-parameter Mixture-of-Experts model (3B active per token) from Kakao (Kanana) (Rest of Asia), released 2026-01-14 under the Kanana License license. It scores 50.7 on the Small LM Index, #29 of 34 medium models. It needs about 20.2 GB of memory at 4-bit with an 8K-token context.
Benchmarks
Each bar shows this model's score. The tick marks the best medium model on that benchmark. Sources: 6 reported by Kakao (Kanana). How we label sources.
Not reported: estimated
What the index model expects from this model's other scores. Not a measurement.
Specifications
- Lab
- Kakao (Kanana) (KR · Rest of Asia)
- Architecture
- Mixture-of-Experts
- Total parameters
- 30.7B
- Active parameters
- 3B
- Size class
- Medium
- Context window
- 32K tokens
- Input modalities
- text
- Reasoning
- Reasoning (thinking)
- License
- Kanana License
- Openness
- Open weights
- Memory to run
- ~20.2 GB at 4-bit, 8K context (more sizes)
- Hugging Face
- 1.3K downloads (30 days) · 63 likes
Compare with
- Kanana-2 30B-A3B Thinking vs Muse Glimmer 30B (29.6B, index 72.6)
- Kanana-2 30B-A3B Thinking vs Gemma 4 31B (30.7B, index 70.1)
- Kanana-2 30B-A3B Thinking vs Granite 4.2 30B (30B, index 67.0)
- Kanana-2 30B-A3B Thinking vs GLM-4.7-Flash (31B, index 66.1)
- Kanana-2 30B-A3B Thinking vs Qwen3 30B-A3B Thinking 2507 (30.5B, index 59.3)
Sources
Something wrong here? Report it.
Run it locally
No GGUF build from a well-known publisher yet.
vllm serve kakaocorp/kanana-2-30b-a3b-thinking-2601Memory needed: weights, context cache and runtime buffers.
| Quantization | 4K context | 32K context |
|---|---|---|
| 16-bit F16 | 63 GB | 64 GB |
| 8-bit Q8_0 | 34 GB | 36 GB |
| 4-bit Q4_K_M | 20 GB | 21 GB |