Unverified50% confidenceFactExact time
ERNIE 5.5 Flash Base has been open-sourced on Hugging Face, supports vLLM and SGLang inference deployment, and is offered in FP8 and BF16 formats, with no NVFP4 format available for now.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
11/5/2026
First Seen
Valid until: 2/3/2027
Sources
Related Entities
Related Claims
UnverifiedThis FlashMLA release is a breaking change: it removes support for the Hopper architecture and V3, V3.2, V4.0, and changes the KV Cache format for FP8 and FP4, making it incompatible with previous versions.70% similarUnverifiedDeepSeek updated V4.1 Flash to the 2017 version and debuted the Reflection feature.69% similarUnverifiedBeam is offered in three formats—BF16, FP8, and NVFP4—supports deployment via Amazon SageMaker JumpStart, and can also run on Kubernetes clusters running vLLM or SGLang.68% similarUnverifiedUnsloth has released a GGUF quantized version of Qwen3 Flash Next, available for download through the Model Hub of the Unsloth desktop application.67% similarUnverifiedIn 5.18.0, GGUF integration added support for MoE, the compressed-tensors FP8 kernel can now run on MPS (torch ≥ 2.15), an OpenVINO HF exporter was introduced, and Flash Attention support was added for XPU and MUSA platforms.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1140871API
curl https://kongchang.com/api/v1/knowledge/claims/1140871MCP
get_claim(id=1140871)