Unverified50% confidenceBenchmarkExact time
The Q4_K_M/GGUF version of Qwen 27B decodes at approximately 10.7 tok/s on the same machine, roughly 1/3.8 of the MLX version.
1
Sources
50%
Confidence
Long-term
Relevance
10/2/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedDeploying Qwen3 27B with 16 GB of VRAM limits the choice to the Q3_K_XL GGUF quantization version, which is the highest precision that 16 GB of VRAM can accommodate.77% similarUnverifiedQwen3 27B's local inference speed is approximately 10 tokens per second.77% similarVerifiedQwen2.5-Coder-32B running with Q4 quantization on an RTX 4090 can achieve an inference speed of 40–60 tokens per second.76% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/968804API
curl https://kongchang.com/api/v1/knowledge/claims/968804MCP
get_claim(id=968804)