Spark-X2.6-4B: Advancing the Frontier of On-Device Agents for Real-World Tasks (Hugging Face) 22 Oct 2026
- BFCL-V4 71.8
- τ²-bench 86.9
- τ³-bench 34.7
- VitaBench2.0 30.8
- MCP-Atlas 62.4
- EnterpriseOps-Gym 33.6
- ClawEval-EN 69.9
- BrowseComp 53.1
- SWebBench (in-house) 71.7
- PAgentBench (in-house) 74.1
- SWE Pro 54.3
- SWE Verified 62.8
- SWE Multilingual 64.7
- SciCode 36.8
- Gaokao 2026 134.9
- AIME 2026 91.2
- HMMT Feb 26 81.3
- IMO-AnswerBench 75.2
- AA-LCRv1.1 67.3
- LongBench v2 53.3
- Mean of Graphwalks bfs 256k f1 92.2 / Graphwalks bfs 1m f1 34.2 63.2
- Mean of Openai_MRCR v2 256k-512k 65.6 / 512k-1M 26.0 45.8
- IFEval 93.6
- IFBench 75.8
- Best of HLE(w/o tools) 17.9 / HLE(w tools) 36.9 36.9
- GPQA Diamond 77.1