Aya Expanse 32B

Overview

The 32B sibling of Aya Expanse from Cohere For AI, delivering a 25% gain on low-resource languages and 89.9% win rate on Dolly vs Mixtral 8x22B. CC-BY-NC.

When to pick this model

• You're doing high-quality multilingual research at the 30B tier
• You need top-tier low-resource language performance
• You're comparing against Mixtral 8x22B on multilingual benchmarks
• Non-commercial use is acceptable for your project

VRAM requirements by quantization

Quantization	VRAM required
Q4_K_M (recommended)	19 GB
Q5_K_M	23 GB
Q8_0	35 GB
FP16 (no quantization)	64 GB

VRAM figures include model weights plus a typical 8k KV cache and ~600 MB runtime overhead (Ollama / llama.cpp baseline). Add headroom for higher context lengths.

Published benchmark scores

Benchmark	Score
Dolly (vs Mixtral 8x22B)	89.9

Scores published by the model author or aggregated from public leaderboards. Re-measured monthly by our editorial team.

Strengths

• 25% improvement on low-resource languages vs peers
• 23 language coverage
• 89.9% win rate on Dolly vs Mixtral 8x22B
• Strong general performance for its size

Limitations

• CC-BY-NC 4.0 — no commercial use
• Only 8K context window
• Newer Qwen 3 models close much of the gap with permissive licenses

Architecture & training

Architecture: Dense (Command R base) · 23 languages

Training: Multilingual fine-tune of the Command backbone.

Quick start

ollama run aya-expanse:32b

Or use the open-source MCP server to query this model from Claude Desktop, Cursor, or any MCP-compatible client.