Community
Unsloth benchmarks dispute ISTA's GSQ-RCO quantized DeepSeek V4.1 Flash claims
Unsloth reports its Dynamic 3.1 GGUFs beat GSQ-RCO by 10 to 20 percent at 512K context. It measured GSQ-RCO IQ1_M at 97.24 percent recovery, below the claimed 98.7 percent, and calls the IQ2_XXS claim nearly impossible.
Every outlet that ran it 1 article
r/LocalLLaMACommunityLead
Unsloth vs GSQ-RCO - 512K context DeepSeek-V4.1 GGUF benchmarks Terminal Bench