DeepSeek-V4-Flash-DSpark published on Hugging Face under MIT license
DeepSeek published the DeepSeek-V4-Flash-DSpark model on Hugging Face, tagged as a text-generation model compatible with transformers, FP8 quantised, and inference-endpoint compatible. The repository lists an MIT license and references paper arxiv:2606.19348. No capability or benchmark results beyond the metadata tags are available.
PUBLISHED2026-06-27
OBSERVED2026-08-11
AGE2mo
SOURCES1
- Pipeline tag: text-generation
- Library: transformers
- License: MIT
- Quantisation: FP8 (8-bit)
- Endpoints compatible
- Paper: arxiv:2606.19348
- Tags: deepseek_v4, region:us
COMMUNITY
No curated reactions recorded for this event. Facts and takes are kept in separate layers — community context is added by hand, never blended into the record above.