Spark-X2.6-1.7B: less than half the parameters of our 4B, close on two benchmarks:
SWE Verified: 59.8 vs 62.8
τ²-bench: 85.3 vs 86.9
Both at high thinking.
Try it & Like on HF!
@huggingface@Ollama@ModelScope2022
🚀 SparkLLM open-sources Spark X2.5-4B & X2.5-1.7B — the only on-device models with native context up to 1,000,000 tokens. Hybrid-attention architecture, tuned for agents, code, math & instruction following. 🧵👇