Leaderboard / OpenBMB (MiniCPM)
AgentCPM-Explore 4B
OpenBMB (MiniCPM) · AgentCPM · released 2026-01-11 · data checked 2026-11-03
AgentCPM-Explore 4B is a 4B-parameter dense model from OpenBMB (MiniCPM) (China), released 2026-01-11 under the Apache-2.0 license. It doesn't have enough reported benchmarks for an index score yet. It needs about 4.2 GB of memory at 4-bit with an 8K-token context.
Deep-research agent model: reports GAIA, BrowseComp and tool-using HLE, which aren't comparable with the no-tools benchmarks used here.
Benchmarks
Each bar shows this model's score. The tick marks the best tiny model on that benchmark.
No benchmark results for this release yet. Know of one? Send it.
Specifications
- Lab
- OpenBMB (MiniCPM) (CN · China)
- Based on
- Qwen3 4B Thinking 2507 (fine-tune or distillation)
- Architecture
- Dense transformer
- Total parameters
- 4B
- Active parameters
- All (dense)
- Size class
- Tiny
- Context window
- 262K tokens
- Input modalities
- text
- Reasoning
- Reasoning (thinking)
- License
- Apache-2.0
- Openness
- Open weights
- Memory to run
- ~4.2 GB at 4-bit, 8K context (more sizes)
- Hugging Face
- 1.5K downloads (30 days) · 418 likes
Compare with
- AgentCPM-Explore 4B vs Qwen3.5 4B (4B, index 57.8)
- AgentCPM-Explore 4B vs Qwen3 4B Thinking 2507 (4B, index 55.7)
- AgentCPM-Explore 4B vs Nemotron Labs Dory 1.0 4B (4B, index 60.0)
Sources
Something wrong here? Report it.
Run it locally
No GGUF build from a well-known publisher yet.
vllm serve openbmb/AgentCPM-ExploreMemory needed: weights, context cache and runtime buffers.
| Quantization | 4K context | 32K context | 256K context |
|---|---|---|---|
| 16-bit F16 | 9.2 GB | 13 GB | 47 GB |
| 8-bit Q8_0 | 5.5 GB | 9.7 GB | 43 GB |
| 4-bit Q4_K_M | 3.6 GB | 7.8 GB | 42 GB |