Back to Ecosystem
Live on Hugging Face

AESIR Coder 7B

A high-performance, reasoning-dense language model developed by AESIR Unlimited. By merging the world-class syntax precision of Qwen2.5-Coder with the rigorous Chain-of-Thought logic of DeepSeek-R1, we engineered a "pocket-sized architect" capable of handling complex software engineering and Web3 tasks on consumer-grade hardware.

View on Hugging Face

Architecture & Methodology

AESIR Coder 7B was engineered using the TIES (Trimming, Electing, and Merging) method via mergekit. This approach resolves weight conflicts between its two parent models — ensuring the reasoning logic of DeepSeek doesn't break the specific coding syntax of Qwen.

Base Engine
Qwen2.5-Coder-7B-Instruct
Optimized for 90+ programming languages
Reasoning Layer
DeepSeek-R1-Distill-Qwen-7B
Trained on massive reasoning traces
Merge Method
TIES via mergekit
Conflict-resolved weight merging
Hardware Profile
8GB–16GB RAM
Local-first, consumer-grade

Key Agentic Capabilities

Chain-of-Thought Auditing

Unlike standard models that just write code, AESIR Coder "thinks" through the logic. Ideal for identifying logic flaws in smart contracts and complex Python agentic systems.

Web3 Native

Deep knowledge of Solidity, Vyper, and the TON/Solana ecosystems, combined with the ability to reason about decentralized state-machines and on-chain protocols.

Structured Intelligence

Highly stable at generating JSON schemas and YAML configurations required for AESIR Protocol agentic handshakes and multi-agent workflow definitions.

Local Execution Standards

To run AESIR Coder 7B on your local machine, we recommend LM Studio or Ollama with a Q4_K_M GGUF quantization for maximum compatibility.

RAM Availability
8GB RAM
Recommended
Q4_K_M
Fast, High Quality
RAM Availability
12GB+ RAM
Recommended
Q6_K or Q8_0
Near-Lossless

Developed by AESIR Unlimited as part of the AESIR Unlimited mission to bridge the gap between AI and Decentralized Infrastructure — making sovereign, open-source AI accessible on consumer hardware.