[Feature]: Support for iFlytek's Newly Open-Sourced Edge-Side Models (Spark X2.5-4B & 1.7B)
Maintainers usually reply within 4 days
Nobody has claimed this yet.
Assessment
- Difficulty
- 5/5
- Estimated time
- Over a week
- Newbie friendliness
- 35/100
- Issue type
- Feature
- Clarity
- Needs clarification
- Activity status
- Active
- Domain
- ai, embedded-iot
Research direction
The issue provides links to the Spark X2.5-4B and 1.7B models but names no files or tests. Start by evaluating those models against FastFlowLM's supported-model scope; done should be a clear decision on whether support is feasible and what follow-up is required.
Written by the indexing model from the issue text.
Description
Suggestion Description
Background
I'm a long-time user of iFlytek's products, particularly their Smart Office Book X5. The device is powered by iFlytek's edge-side models, which have been specifically fine-tuned for meeting transcription and light office tasks. In my experience, the performance on these specialized tasks is quite impressive.
Why These Models Matter
These models, being compact and efficient, appear to be well-suited for always-on deployment on NPUs like ours, which could enable seamless, day-to-day office assistance directly on the device. The X5 itself leverages a powerful NPU, demonstrating the viability of this approach.
The Opportunity
I'm excited to see that iFlytek has just open-sourced two of these edge-side models today (September 1, 2026):
A key feature is their native support for a 1 million token context window, which is quite advanced for edge-side models.
Given their proven performance in office scenarios, optimization for edge deployment, and now open-source availability, I believe these models would be a valuable addition to [Your Inference Engine's Name]'s supported model list. They could open up new possibilities for on-device AI assistance.
Would the team be interested in evaluating these models for future support?
Thanks for considering!
Operating System
Windows11
GPU
Strix Halo AI MAX 395
ROCm Component
10.0
- Dominant language
- C++
- Stars
- 1.9k
- Forks
- 158
- Avg merge
- 4d 16h
- Merged PRs (30d)
- 14
Getting set up
- Ships a Dockerfile or Docker Compose file
- No pull request template
- Read the contributing guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
More from ROCm/FastFlowLM
-
Difficulty 2/5 1-3 hours Newbie friendliness 78/100
ROCm/FastFlowLM#757 ·
Maintainers usually reply within 4 days
-
No valid checkpoint to restore + Max length reached on multi-turn tool calls (Qwen3.6-MoE, NPU)Open
Difficulty 2/5 1-3 hours Newbie friendliness 82/100
ROCm/FastFlowLM#744 · 2 comments ·
Maintainers usually reply within 4 days
-
Truncated tool call is emitted as a complete one (unterminated heredoc reaches the client)Possibly taken @Javinator9889 claimed this 19 days ago. Open
Difficulty 2/5 1-3 hours Newbie friendliness 78/100
ROCm/FastFlowLM#741 · 1 comment ·
Maintainers usually reply within 4 days
-
Download timeoutPossibly taken @stdevPavelmc claimed this 147 days ago. Open
Difficulty 2/5 1-3 hours Newbie friendliness 74/100
ROCm/FastFlowLM#520 · 7 comments ·
Maintainers usually reply within 4 days
-
Difficulty 2/5 1-3 hours Newbie friendliness 68/100
ROCm/FastFlowLM#434 · 1 comment ·
Maintainers usually reply within 4 days
Similar issues
-
8-membered-ring atrop stereo lost in 2026.09.1Possibly taken A pull request linked to this issue is open or already merged. Openbug
Difficulty 2/5 Half a day Newbie friendliness 86/100
Maintainers usually reply within 2 days
-
thread safetyOpen1.0
Difficulty 1/5 Under an hour Newbie friendliness 72/100
Maintainers usually reply within 1 day
-
libasr headers?Open
Difficulty 2/5 1-3 hours Newbie friendliness 66/100
Maintainers usually reply within 1 day
-
Type: Bug :bug:
Difficulty 2/5 1-3 hours Newbie friendliness 83/100
-
llvm-trunk
Difficulty 1/5 Under an hour Newbie friendliness 88/100
Maintainers usually reply within 1 day