# SGLang adds Day-0 support for Spark-X2.6-1.7B, a compact on-device model

> Original title: SGLang has Day-0 support for Spark-X2.6-1.7B from @SparkLLM, a compact on-device model that stays close to Spark-X2.6-4B with less than h...

- Source: SGLang
- Published: 2026-10-28T06:46:23.000Z
- Bad Influence: https://badinfluence.news/items/a0dx6the2448owbs6pzv7s85h
- Original: https://x.com/sgl_project/status/2115334343870787975

## Summary

SGLang announces Day-0 support for Spark-X2.6-1.7B from SparkLLM, a model with less than half the parameters of Spark-X2.6-4B. The quoted SparkLLM post reports SWE Verified scores of 59.8 versus 62.8 and τ²-bench scores of 85.3 versus 86.9 for the two models at high thinking. The post includes a Docker command to launch the model with SGLang.

## Post

SGLang has Day-0 support for Spark-X2.6-1.7B from @SparkLLM, a compact on-device model that stays close to Spark-X2.6-4B with less than half the parameters.

> Stronger coding, agentic, and long-context
> Efficient on-device inference, same architecture as Spark-X2.5-1.7B
> Controllable Thinking Budget, supports none, low, and high levels

Try it with SGLang!

![](https://badinfluence.news/api/img-proxy?u=https%3A%2F%2Fpbs.twimg.com%2Fmedia%2FHVss3T6awAEf0Bq.jpg&mode=thumb&exp=1793404800&sig=a03c9091c4103547)

## Quoted post @SparkLLM

> 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!
> https://huggingface.co/XHToken/Spark-X2.6-1.7B
> @huggingface @Ollama @ModelScope2022

https://x.com/SparkLLM/status/2115100695775056057