FULL STORY
Unsloth's DeepSeek-V4.1-Flash Quant Release Sparks Benchmark Feud
Unsloth released GGUF quantizations of DeepSeek-V4.1-Flash with leading 1-bit benchmark scores, then became embroiled in an escalating benchmark dispute with ISTA (GSQ-RCO), each side accusing the other of withholding testing details.
2026-11-02 ~ 2026-11-03 · 2 episodes · 6 posts
Episode 1 · Unsloth Releases DeepSeek-V4.1-Flash GGUF, 1-bit Leads Quantization Benchmarks (2026-11-02, 3 posts)
Unsloth released GGUF quantizations of DeepSeek-V4.1-Flash, benchmarking them with 35.1B tokens at 512K context; its 1-bit version leads other quantizations by over 10% on DeepSWE, while some GSQ-RCO recovery rates were found inflated.
- Unsloth's 1-bit DeepSeek-V4.1-Flash GGUF beats other quants by 10%+ on agentic benchmarks — solyarisoftware · 2026-11-02
- Unsloth's 35B-token benchmarks challenge GSQ-RCO's DeepSeek quant recovery claims — danielhanchen · 2026-11-02
- Unsloth quants top agentic benchmarks: 1-bit GGUF scores 68.1 on DeepSWE, 10% ahead — danielhanchen · 2026-11-03
Episode 2 · Unsloth and GSQ-RCO clash over quantization benchmarks (2026-11-03, 3 posts)
The quantization benchmark dispute between GSQ-RCO and Unsloth is escalating. Unsloth's Daniel Hanchen says retests on 27 billion tokens still show a 10-20% lead and accuses GSQ-RCO of refusing to disclose methods, while Reddit users question why GSQ-RCO hasn't published KLD metrics.
- Unsloth rebuts GSQ-RCO with 35B-token benchmarks; rival still won't share methods — danielhanchen · 2026-11-03
- Unsloth responds to quant benchmark dispute: reran 128K LCB 6-12x, own quants scored worse too — danielhanchen · 2026-11-03
- ISTA (GSQ-RCO) vs Unsloth quant beef heats up as users ask why ISTA won't publish KLD — Queasy_Back2588 · 2026-11-03