Confirmed
UkisAI releases Swift GLM-5.3-Flash and Swift Ornith 9B with reduced reasoning tokens
UkisAI reports 61.4% fewer thinking tokens on MMLU-Pro for Swift GLM-5.3-Flash with a 0.46% average accuracy loss, and 33.5% fewer thinking tokens for Swift Ornith 9B with no accuracy loss. Both come with GGUF quants, and the company says a GSQ-RCO Q3 quant will follow.