You need to agree to share your contact information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

Eyght v18

Eyght v18 is a fine-tune of Qwen/Qwen2-7B tuned for coding, math, reasoning, and general knowledge across Python and C/C++, with debugging ability. Packaged as a Q4_K_M GGUF for Ollama.

Built, trained, and owned by Eyght. Free Hugging Face model repository.

Model details


Base model	`Qwen/Qwen2-7B`
Architecture	Qwen2ForCausalLM (decoder-only)
Parameters	~7.6 B
Format	GGUF (Q4_K_M) + LoRA adapter
Runtime	Ollama / llama.cpp
License	Apache-2.0 (fine-tune)

How to use

ollama run eyght-v18

Benchmark Results

Evaluated on the Eyght Arena benchmark (20 problems across 3 categories):

Category	Problems	Passed	Pass Rate
Coding (algorithm implementation)	10	5	50%
Bug Fix (SWE-bench-style code repair)	5	1	20%
Tool Use (MCP-style function calling)	5	2	40%
Overall	20	8	55%

Project Janus

This model is designed to operate within Project Janus, a dual-loop cognitive architecture:

Perceive -> Working Memory -> Internal Critic -> Action -> Consolidation

Internal Critic evaluates outputs before execution
Episodic Store remembers past interactions
Core Value Axioms enforce safety rules

License

Apache-2.0 - Built by Eyght.

Downloads last month: -; Downloads are not tracked for this model. How to track

Model tree for Eyght/eyght-v18

Base model

Qwen/Qwen2-7B

Adapter

(225)

this model