SparkLLM

Leading general-purpose foundation models & professional AI agents. The proprietary Spark model family, with AStudio & Loomy harnesses. Open source.

[SparkLLM @SparkLLM]

100K+ downloads in 10 days! 💜

Spark X2.5 has passed 100,000 downloads across its model family on Hugging Face.

Thank you for downloading, testing, and sharing.

Try the 1.7B and 4B models: https://t.co/D2iMd5hnMM

[SparkLLM @SparkLLM]

Share your real, reproducible findings. Deadline: Sep 13, 2026, 24:00 Beijing time (UTC+8). One winner, USD 100. Full FAQ: https://t.co/wVYAz9Mi5Q #HERHackAstron

[SparkLLM @SparkLLM]

Right answer, wrong explanation? FU-max-boop's Spark-X2.5-1.7B case shows why reasoning deserves a closer look. Watch the short, then read the HER Hack-Astron #6 FAQ below. https://t.co/cNBEgNv48y #SparkX25 #HERHackAstron

[SparkLLM @SparkLLM]

FAQ 9-11: No upfront fee is required. Provide sensitive identity and payment information only privately, if needed after winning. No live presentation is currently required.

[SparkLLM @SparkLLM]

6/6 International winners may use available payout methods such as PayPal or bank transfer. Mainland China winners may coordinate RMB payout through an available domestic method.
https://t.co/CKKxc3EZNi

[SparkLLM @SparkLLM]

1/6 Right answer, arithmetic slip? Ykmmz’s Spark-X2.5-1.7B entry reports 160 runs on an Apple M3 with 16 GB. Watch case #20, then explore the challenge FAQ below. https://t.co/GhVTATgtAU #SparkX25 #HERHackAstron

[SparkLLM @SparkLLM]

5/6 Actual, reproducible model runs are required. Submit via Hugging Face Discussion + GitHub Issue. No upfront fee; no live presentation currently required. Sensitive ID/payment info: privately, only if needed after winning.
https://t.co/CKKxc3EZNi

[SparkLLM @SparkLLM]

Separate tests with an Unsloth-built llama.cpp backend also pass this example. Try compatible GGUF runtimes and your own tasks. Share feedback, experiences and creations: https://t.co/x9AlVyYoys #LocalAI #OpenSource

[SparkLLM @SparkLLM]

Spark-X2.5 on Unsloth + LM Studio! Watch a real local GGUF demo: Chinese and English both solve 3(x-2)+4=19 and verify x=7. https://t.co/RmkDmJpf0m #SparkX25 #Unsloth #LMStudio

[SparkLLM @SparkLLM]

Official Spark-X2.5-1.7B Q4_K_M, real LM Studio recording. The model expands the brackets, combines terms and checks 3(7-2)+4=19. Footage is labelled 3x speed; full original-speed recordings are retained.

[SparkLLM @SparkLLM]

Bring your own reproducible math test. Publish the full evaluation in the model’s HF Discussions, then reply to Issue #9 with the link. Sep 13, 2026, 24:00 Beijing time; one winner, USD 100. Full rules: https://t.co/CKKxc3EZNi #HERHackAstron

[SparkLLM @SparkLLM]

Right math, rejected answer? Anmolspace’s Spark-X2.5-1.7B case found a correct 5/33 answer that failed strict formatting. Watch the short: https://t.co/m26W5s91Jb #SparkX25 #HERHackAstron

[SparkLLM @SparkLLM]

We recalculated saved-output scores without rerunning inference. The source reviewed disclosed AI-led experiments and unconfirmed complete human review. Organizers decide eligibility. https://t.co/Xgg8JQamVG