Tasks and inputs
Spark X2.6 is listed for Language & text, Text → Text. It accepts text and produces text. Use the publisher’s input format and preprocessing for your chosen deployment.
Spark X2.6 is iFLYTEK's 293B mixture-of-experts language model family with integrated device and cloud variants, unveiled in October 2026.
The specification score summarizes recorded features and limits. It is not a benchmark result.
Recorded modality support shows what this model can accept and produce. An undocumented modality is shown as unknown rather than assumed to be unsupported.
Accepts
Produces
Not much is public yet about this model, and that shapes how to approach it. iFLYTEK unveiled the Spark X2.6 family at its 1024 Global Developer Festival in Beijing, which opened on 22 October 2026. The company describes a 293-billion-parameter mixture-of-experts cloud model with integrated on-device and cloud capabilities.
That is about all iFLYTEK's own announcement confirms.
| Item | Status |
|---|---|
| Unveiled | 22 October 2026, iFLYTEK 1024 Global Developer Festival |
| Architecture | Mixture-of-experts, 293B total (vendor-stated) |
| Active parameters | About 30B reported; not in the vendor release |
| Context | 1M tokens reported; not in the vendor release |
| Open source | Reported; no licence or download location confirmed |
| Pricing and API | Not stated |
Press coverage adds that the model was built and served on fully domestic compute, and that iFLYTEK claims gains in coding and agentic tasks. Those claims are the company's and have not been independently tested.
The predecessor, Spark X2.5, arrived in September 2026 with a 293B-A30B design and a 256K context. If the reported 1M window for X2.6 holds, the main step is context length rather than size.
Until those details firm up, better-documented open MoE options include Baidu's ERNIE 5.5 Flash line and Meituan's LongCat-2.0; you can line any of them up side by side in the model comparison tool. ### Why the gaps matter
A missing context figure or licence is not a footnote for a model of this size. A 293B MoE needs a multi-GPU node to self-host, and the business case changes completely depending on whether the weights are free to use commercially, whether the on-device variant is a separate small model, and whether a hosted API exists outside mainland China. Those three answers decide whether Spark X2.6 belongs on a shortlist at all.
We will expand this page once iFLYTEK publishes model cards. Meanwhile, the text-to-text directory lists comparable models.
Compare the model on the work you need to complete.
Spark X2.6 is listed for Language & text, Text → Text. It accepts text and produces text. Use the publisher’s input format and preprocessing for your chosen deployment.
Measure a completed task using the prompts, documents and tools your application will use. Include reasoning and retries in the result.
Use the model identifier and account requirements for the selected API route. Confirm its request limits and output format with one complete task before connecting a larger workflow.
These records share a directory category. Check the source card, output type and deployment requirements before treating a model as a replacement.
baidu
ERNIE 5.5 Flash Preview is Baidu's first post-trained 280B-A14B text model, with a 1M window, aimed at game dev, 3D scripting and knowledge work.
Meta
Muse Ember is Meta's smallest, cheapest Muse model: a 1M-context multimodal worker for tool calls, coding subtasks, routing and extraction at volume.
Meta
Muse Spark Open is Meta's open-weight 816B mixture-of-experts model (42B active) with 1M context, built for coding and agents you can self-host.
ByteDance Seed
Seed 2.1 Lite is a lightweight Doubao Seed 2.1 model listed on Volcengine Ark with a 1M context, 256K output, audio and image input and tool calls.