LinkkodeeAnnouncement
Fine-tuned models at the 27B scale frequently drop features like multimodal vision or multi-token prediction during adaptation. Qwythos-27B retains the entire architecture intact and fits comfortably onto consumer 24GB GPUs using the Q4_K_M quantization.
Empero takes a transparent release approach, prioritising measured target-format acquisition and local execution over broad, unvalidated benchmark claims. Retaining Q8_0 precision for the sensitive SSM state path within lower K-quants ensures structural integrity for linear attention at a minimal size trade-off. The unaligned, uncensored nature makes it capable for research and code execution, though production deployments will require external guardrails.
Summary generated by AI (Gemini 3.8 Flash, 27 September 2026). It may contain errors – the original source is authoritative.
Similar in content, found by the AI search.
Link:Empero AI
Empero AI has released Qwythos-27B-v1, an open-weight reasoning model based on Qwen3.5-27B. The checkpoint preserves multi-token prediction (MTP), vision capabilities, and a 1M-token context window while specialising in terminal and agentic workflows.
Link:empero.org
Empero, an independent AI research lab based in Germany, develops efficient open-weight language models and developer tools for local hardware. Releases include Qwen3.8 distillations, the Qwythos model family, and the Abacus terminal coding agent.
Link:Unsloth AI
Unsloth AI has published Qwen3.6-27B-MTP-GGUF, providing quantized versions of the first open-weight model in the Qwen3.6 family. The 27-billion parameter multimodal model features Multi-Token Prediction (MTP) for accelerated inference alongside upgrades to agentic coding and tool calling.
Link:Unsloth AI
Unsloth AI has published GGUF quantizations for the Qwen3.8-27B model under an Apache-2.0 licence. Utilizing their new Dynamic V3.0 quantization format, Unsloth claims over 10% higher top-1 accuracy at identical sizes compared to alternative quants. Qwen3.8-27B is a 27-billion parameter dense vision-language model featuring native text, image, and video comprehension alongside granular thinking controls.
Link:Qwen
The Qwen team has released Qwen3.8-27B, a compact, dense vision-language model under the Apache-2.0 licence. The 27-billion-parameter model pairs Gated DeltaNet with conventional gated attention, handles images and videos natively, and introduces flexible reasoning controls.
Link:PocketAiHub
PocketAiHub has released an unofficial MLX derivative family of Qwen3.8-27B with suppressed refusal behaviour ('abliteration'). The consolidated repository provides quantized builds ranging from experimental 2-bit AWQ to full BF16 optimised for Apple Silicon.
Parts of this content were created with the assistance of AI.
