Spark-X2.6-1.7B launches with 1M context and Apache 2.0 for on-device agents
AdinaYakup · x · 2026-10-27
Following its 4B model, @SparkLLM released the smaller Spark-X2.6-1.7B, targeting on-device agents. Key features: three thinking modes (none/low/high), up to 1M context, Apache 2.0 license, and multiple quantized formats (GGUF, INT4, INT8, FP8). Reported tests claim leading decoding throughput among similarly sized models. The 4B version is already popular on Hugging Face (6k+ downloads for the main model, 100k+ for GGUF), while the 1.7B variant just landed.
Related event: iFlytek Open-Sources Spark-X2.6-1.7B with 1M Context(3 posts)→
More from Infra
- General Compute raises $100M Series A to bet on fast inference demand — _akhaliq · 2026-10-28
- Microsoft's custom Maia chips to make up only ~10% of its AI chip spend by early 2027, says Evercore — Beth_Kindig · 2026-10-28
- YC F26 startup Applied Kin quotes pad-mount transformers in 7 weeks as AI-native manufacturer — ycombinator · 2026-10-28
- Cowboy Space unveils 1MW orbital GPU satellite: 5 GPU racks, 132m span, laser links — BenBajarin · 2026-10-28
- Open-source Zilla adds LLM gateway support: preserving API fidelity across providers — jkriket · 2026-10-28
- Together breaks down the MIGHT stack: composable open-source AI from models to MCP tools — togethercompute · 2026-10-28