January 26, 2026
Charts vs comments
Qwen3-Max-Thinking
Qwen’s “super thinker” debuts, but fans want proof, prices, and a Hugging Face link
TLDR: Qwen3-Max-Thinking claims top-tier reasoning and on-demand tools, even beating Gemini in some tests. Commenters care more about missing open release, unclear pricing, China-only discounts, and whether it actually wins at coding—lots of buzz, not enough answers.
Qwen just dropped Qwen3-Max-Thinking, a big-brain AI boasting charts that claim it rivals GPT, Claude, and Gemini—and even beats Gemini on some reasoning tests thanks to fancy “think harder at runtime” tricks. It also promises on-demand tools like web search and a code interpreter at chat.qwen.ai. Sounds epic… on paper.
But the comments turned into a spectator sport. The loudest take: it still doesn’t top Opus 4.5 in “agentic coding” (bots that write and fix code), so devs like throwaw12 are groaning they’ll “wait more.” Another camp is side-eyeing the rollout: siliconc0w can’t find a Hugging Face link, sparking “is this still open or just a cloud thing?” debates. Meanwhile, the money talk is messy—Mashimo asks if it’s subscription or pay-per-token, and isusmelj points out prices look cheaper in mainland China, which triggered geo-pricing outrage and armchair economics. Then arendtio drops the mic: “How big is it?”—because Qwen brags about scaling up, but no model size is listed.
Cue the memes: “Schrödinger’s model—measured, but unseen,” “Benchmark billionaire, budget mystery,” and “Think harder, charge softer.” The vibe? Hype for the brainpower, chaos over the download, dollars, and details. Read the flexes on the blog and the pricing maze on Alibaba Cloud.
Key Points
- •Qwen launched Qwen3-Max-Thinking, a flagship reasoning model with scaled parameters and extensive reinforcement learning.
- •The model introduces adaptive tool-use (on-demand retrieval and code interpreter) and advanced test-time scaling to boost reasoning.
- •Across 19 benchmarks, Qwen reports competitive performance with top models and claims to surpass Gemini 3 Pro on key reasoning tests.
- •A detailed table lists scores in knowledge, STEM, reasoning, agentic coding/search, instruction following, tool use, and long-context handling.
- •Qwen3-Max-Thinking is available via Qwen Chat and through the Alibaba Cloud Model Studio API.