| Model | Score | Status | Released | Price in/out | Context | |
|---|---|---|---|---|---|---|
| Qwen3-235B-A22B frontier-adjacent reasoning at open-weight prices | 8.5 | GA | 2025-04-28 | $0.20 / $0.60 | 131K | Review → |
| Qwen3-32B best single-GPU open-weight model in production | 8.4 | GA | 2025-04-28 | $0.08 / $0.28 | 131K | Review → |
| Qwen3-14B best open-weight model in the 13-15B band | 7.8 | GA | 2025-04-28 | $0.06 / $0.20 | 131K | Review → |
Qwen3-235B-A22B is Alibaba's open-weight flagship, shipped 2025-04-29 under Apache 2.0. It is a Mixture-of-Experts model — 235B total parameters but only 22B activated per token (the "A22B" suffix) — that brings frontier-adjacent math, code, and reasoning into the open-weight tier with a hybrid "thinking / non-thinking" mode toggle in a single set of weights. For a buyer, the one-sentence pitch is: DeepSeek-R1-class reasoning and broad multilingual quality, redistributable under Apache 2.0, at roughly 1/30th the per-token cost of a Western frontier model. - Provider: Alibaba Cloud (Qwen Team) - Released: 2025-04-29 (GA) - Tier: Large open-weight MoE flagship - Context: 131,072 tokens - Max output: 32,768 tokens - Modalities: text in, text out - Knowledge cutoff: approx. 2024-10 - Headline price: $0.20 in / $0.60 out per 1M tokens (Together AI)
Full review →