Unsloth and GSQ-RCO dispute benchmark results for DeepSeek-V4.1-Flash GGUF quants

1. 11:52 AM
   Daniel Han@danielhanchenX

   Unsloth publishes benchmark methods amid dispute with GSQ-RCO

   AIUnsloth co-founder Daniel Han says the team asked GSQ-RCO for more detail on how its benchmarks were made, and that Unsloth has published its own methods publicly while GSQ-RCO has not. Han says Unsloth ran DeepSWE, Terminal Bench, Automation Bench, HLE, and SciCode at 512K context on its DeepSeek UD3.1 GGUFs, reporting it is more than 10-20% better than GSQ-RCO. He adds that the team used 35.1B tokens of benchmarking, compared with 1.3B before.

   
2. 10:36 AM
   Daniel Han@danielhanchenX

   Unsloth reports DeepSeek GGUF quants lead on 512K-context agentic benchmarks

   AIUnsloth says its DeepSeek-V4.1-Flash-1017 GGUFs score more than 10% higher than GSQ-RCO on DeepSWE, Terminal Bench, Automation Bench, HLE, and SciCode at 512K context. The 1-bit Unsloth GGUF scores 68.1 on DeepSWE, 72.2 on Terminal Bench, and 52.7 on Automation Bench. Unsloth also re-measured GSQ-RCO's recovery figures, reporting lower results than GSQ-RCO claims for its IQ1_M and IQ2_XXS quants.

3. 9:43 AM
   Unsloth AI@UnslothAIOfficial

   Unsloth's 1-bit DeepSeek-V4.1-Flash GGUF leads on DeepSWE and Terminal-Bench

   AIUnsloth says its 1-bit DeepSeek-V4.1-Flash-1017 GGUF scores 68.1 on DeepSWE, 72.2 on Terminal-Bench, and 52.7 on AutomationBench at 512K context. The company says this is more than 10% ahead of other quantizations it tested. The GGUF is available on Hugging Face.

   

The trend compares only continuously observed participants, so its scope may be smaller than the current heat total. Hover or tap for hourly heat; use arrow keys with a keyboard.