Qwen 3 32B Instruct, Solar Pro 22B, and Yi 1.5 34B Chat all sit in the 22–34B parameter range and each comes from an Asian AI lab targeting different language and deployment niches. Comparing them is largely an exercise in context window and licensing trade-offs. Qwen 3 32B Instruct from Alibaba is the broadest-purpose model here: 32B parameters, 131K context, strong multilingual performance across CJK languages and Arabic, and benchmark scores landing at roughly 85 percent of Qwen 3 72B at about half the inference cost. The Qwen license permits commercial use, and provider coverage across Together, Fireworks, Groq, and DeepInfra makes it easy to benchmark. Solar Pro 22B from Korean lab Upstage is a 22-billion-parameter dense model released October 2024, targeting single-GPU deployments and Korean-English bilingual document understanding. Fitting on one A100 or H100 under int8 quantization means predictable infrastructure cost. The CC-BY-4.0 license is the most permissive of the three — attribution only, no share-alike, no commercial restrictions. The 4K context ceiling is the main limitation; anything document-length needs aggressive chunking or a different model. Yi 1.5 34B Chat from 01.AI is a 34B bilingual model for English and Chinese, released May 2024 under Apache 2.0. Benchmark scores were competitive at release, but the 4K context window was already a constraint then, and by mid-2026 Qwen 2.5 and DeepSeek models in the same size class have moved clearly ahead. The remaining case for Yi 1.5 34B is teams that need Apache 2.0 licensing with Chinese support and are willing to work around the context limitation. Pick Qwen 3 32B for general multilingual use with long context. Pick Solar Pro 22B for Korean-English bilingual tasks on a single GPU with CC-BY-4.0 terms. Pick Yi 1.5 34B only when Apache 2.0 plus Chinese support are both hard requirements.