DeepSeek model card and historical trend record · First found 2026-09-10
DeepSeek V4.1 Flash: visual input, long context and open weights
DeepSeek V4.1 Flash accepts images and text, returns text, and provides open weights. Its model card explains reductions in the work of processing long inputs and retaining context.
Images and text enter the same model
The official card describes V4.1 Flash as a native multimodal MoE model with a context limit of one million tokens. It processes encoded visual information together with text, allowing questions about screenshots, charts or documents. Its output is text.
The backbone has 552B parameters, with about 8B active per token during prefill and 16B during decoding. Total model size and active parameters measure different things.
Open weights and context storage
The Hugging Face page provides MIT-licensed weights and inference material. DeepSeek reports a global KV cache of 890 bytes per token, about one quarter of V4 Flash’s footprint. That is a storage measure; API charges depend separately on usage and rates.
The September 10 discovery snapshot remains intact: seventh in this site’s filtered selection, trend score 278, 285 likes and a download counter of zero. Those historical discovery fields have not been replaced with today’s numbers.
Important additions and corrections
Dates below mark changes to our coverage.
- Addition ·
Added visual inputs, context, open weights and cache information while retaining the trend snapshot.
Sources and notes
sources中文
Sources and notes
- Read the originalDeepSeek model card and historical trend record
Retained summary and context
Entry numbers describe this filtered collection, not original HF chart positions or global ranks. Trending score, downloads, and likes: attention is not proof of model quality, adoption, benchmark results, or production readiness. Check the official model page and model card.
Sources and evidence · 2 links
Each link has a purpose and scope. Link counts do not establish independent confirmation; direct support applies only to the statements specified below.
This page keeps the original link and a short account. Follow the links below for the publisher’s full explanation.
Source timeline
- · Entry #7 after this HF trend collection's filters, with a trending score of 278; no original HF chart rank is recorded here.
Collection record
First discovered: ; added to the library:
Collection and source publication have distinct dates.
Editorial record
Sources reviewed
· Beijing time (UTC+8)
Added visual inputs, context, open weights and cache information while retaining the trend snapshot. Original event, discovery and collection dates remain.
Sources and full record
This dates our review of the wording and sources, not a project release or product update.
Full record: before, after and sources (JSON)