Qwopus MoE 35B-A3B lands in GGUF

// 45d agoMODEL RELEASE

Qwopus MoE 35B-A3B lands in GGUF

Qwopus MoE 35B-A3B is a GGUF release built on Qwen3.5-35B-A3B and distilled from Claude Opus 4.6 reasoning traces. It targets local developers who want Opus-flavored reasoning on consumer hardware, with quantizations small enough to run on a single high-end GPU.

// ANALYSIS

This is a clever distillation package, not a new frontier base model: it pairs a sparse Qwen MoE backbone with Opus-style reasoning traces, then ships the result in a format local-stack users can actually deploy. The appeal is obvious for agentic coding and research workflows, but the model card also suggests the usual tradeoff between richer visible output and raw speed.

–The MoE backbone keeps compute low at inference time, with roughly 3B active parameters out of 35B total.
–GGUF quantizations make it practical for local use, including a 4-bit build sized for around 24GB VRAM.
–The model card shows strengths in coding, planning, and research-style output, but also some reasoning and latency regressions versus the earlier Opus-distilled variant.
–This is best read as an opinionated local reasoning model for power users, not a universal replacement for the underlying closed models.

// TAGS

qwopusllmreasoningopen-sourceopen-weightsself-hostedinference

DISCOVERED

45d ago

2026-04-29

PUBLISHED

47d ago

2026-04-27

RELEVANCE

9/ 10

AUTHOR

HuggingModels

// KEEP READING

More AI developer news from the feed

EXPLORE FULL FEED

MODEL23m ago

Zhipu AI drops GLM-5.2 coding model

Zhipu AI has launched GLM-5.2, its new 744-billion parameter Mixture-of-Experts coding model featuring a 1-million-token context window. Currently available to GLM Coding Plan subscribers, Zhipu AI plans to release the model's MIT-licensed weights and API next week.

MODEL35m ago

Claude Fable 5 exposes fragile safety guardrails

Anthropic's Claude Fable 5 was introduced as a safeguarded, public-facing variant of its powerful Mythos-class architecture designed for complex agentic workflows. The release strategy relies on real-time classifiers that reroute sensitive prompts to Claude Opus 4.8, depending on the premise that software guardrails can successfully isolate hazardous capabilities.

NEWS1h ago

Andrej Karpathy shares the "LLM Wiki" design pattern, where an AI agent actively maintains and organizes a user's knowledge base.

The "LLM Wiki" is a design pattern introduced by Andrej Karpathy to address the "memory rot" and organization challenges typical of second-brain systems. Instead of using standard Retrieval-Augmented Generation (RAG) to query a chaotic directory of notes, the pattern proposes a three-layer architecture: raw sources, a synthesized wiki of interlinked markdown files, and instructions for how the AI agent should maintain it. Under this pattern, the LLM acts as an active gardener of the wiki—synthesizing new info, identifying connections, and resolving contradictions—resulting in a compounding knowledge base.

Qwopus MoE 35B-A3B lands in GGUF