Expired50% confidenceFactExact time
The EXL2 format was introduced by the ExllamaV2 project in 2024, supports non-uniform bit-width allocation, and ExllamaV2 currently only supports NVIDIA GPUs (CUDA), without support for Metal or Vulkan.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/16/2026
First Seen
Valid until: 10/14/2026(expired)
Sources
Related Claims
UnverifiedExLlamaV3 是专为消费级 NVIDIA GPU 上高效运行 GGUF/EXL2 等量化格式大语言模型而设计的推理引擎68% similarUnverifiedUnsloth通常需要配合NVIDIA GPU使用，推荐RTX 3090/4090或更高62% similarUnverifiedOpen-source models can be run on consumer-grade GPUs (such as the NVIDIA RTX 4090 with 24GB VRAM) via inference frameworks such as Ollama, llama.cpp, and LM Studio.61% similarUnverifiedExLlama 系列深度针对 NVIDIA CUDA 平台优化，在单用户本地交互场景下往往能榨取更高的单卡性能上限61% similarUnverifiedThe FP8 format was jointly proposed by NVIDIA, ARM, and Intel in 2022.60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/536143API
curl https://kongchang.com/api/v1/knowledge/claims/536143MCP
get_claim(id=536143)