HaiAI123

Curated Global AI Tools Directory

Industry update
New

iFlytek's Spark X2.6 Adds 1M-Token Context, Open-Sources 4B On-Device Model

10/22/2026 — 10/22, 22:19·2 sources (all)·2 reports (all)

Key takeaways

  • At its 2026 1024 Developer Festival main forum on October 22, iFlytek released Spark X2.6, a model that routes tasks between on-device and c…

Background & analysis

On October 22, 2026, iFlytek unveiled its cloud-device integrated Spark X2.6 large model at the main forum of the 2026 Global 1024 Developer Festival, alongside a 4B on-device model, Spark X2.6-4B. A central piece of the announcement was the cloud-device architecture: the system routes each user task to pure on-device execution, pure cloud execution, or a collaborative mode combining both. An architecture diagram shown on stage indicated that the collaborative mode covers two patterns, one in which the cloud guides on-device execution and one in which the two sides split the work, with both leading to task delivery.

During the live demo, the on-device model handled tasks such as email summarization and redaction of sensitive contract information locally, after which the cloud model reviewed and revised the output. In the contract demo interface, the system identified and labeled sensitive fields including company name, legal representative and authorized representative, dates, bank account number, bank branch and commercial amount, counted the categories and their quantities, and replaced some of the sensitive fields. The report noted that real-world recognition accuracy, completeness of redaction and the scope of data transmission would still depend on the specific deployment and usage.

Chairman Liu Qingfeng said the company expects that roughly 60% of future application scenarios can be completed on-device, 30% through cloud-device collaboration, and 10% requiring the cloud model. iFlytek also said the share of tasks the on-device model can complete independently has risen from 12.8% to more than 40%, without disclosing how that figure was measured. The available material stops at this launch and demo; further details of the model's capabilities were not covered.

AI-generated from 1 reports · updated 4 hours ago

Latest turniFlytek has released Spark X2.6, a device-cloud model family whose cloud 293-A30B model supports a million-token context and was upgraded on fully domestic compute, according to Zhidongxi. The 4B on-device model is open sourced, and iFlytek says input cache-miss and output pricing undercuts DeepSeek-V4.1-Flash.

24-hour heatHeat index 176 · peak 190 · 2h ago
24 hours agonow

Reports on this story headlines open the original

Yesterday
  1. iFlytek has released Spark X2.6, a device-cloud model family whose cloud 293-A30B model supports a million-token context and was upgraded on fully domestic compute, according to Zhidongxi. The 4B on-device model is open sourced, and iFlytek says input cache-miss and output pricing undercuts DeepSeek-V4.1-Flash.

    智东西AI score 62
  2. At its 1024 Developer Festival on October 22, iFlytek released Spark X2.6, built around a device-cloud architecture, along with a 4B on-device model. Tasks are routed to on-device, cloud, or hybrid execution; in a demo, the on-device model summarized emails and redacted sensitive contract fields before the cloud model reviewed the result. iFlytek says the share of tasks completed entirely on-device has risen from 12.8% to over 40%, without disclosing how that figure is measured.

    华尔街见闻AI score 56

Other stories people are talking about

How is heat calculated?About the method

Heat counts how many independent sources covered a story in the last 48 hours: one source counts once no matter how many posts it published, decaying with a 24-hour half-life. What ranks first is what many people are talking about.

This page aggregates public feeds. Headlines and summaries are machine-organized and remain the property of the original authors; verify important facts at the source. If you believe a headline or summary infringes your rights, email the contact address in the footer with the page URL and basis of your claim, and we will remove or replace it after review.

Surge
Discussion rising fast
New
First report within 6 hours
Rising
Still gathering discussion

Back to AI industry updates →