ModelsQwen3-Coder-Next: An 80B Coding Agent With 3B Active
Qwen3-Coder-Next pairs 80B total parameters with 3B active and hybrid attention under Apache 2.0. What it is built for, how to run it locally, and where it falls short.
ReadNotes on models, agents, and what inference costs
Practical writing for developers building with large language models — how they work, how to pick one, and how to keep the bill predictable.
Picking between models — and testing them on your own work.
Models
ModelsQwen3-Coder-Plus: Alibaba's Hosted Coder With a 1M Window
Qwen3-Coder-Plus is the hosted, closed version of Qwen3-Coder with a 1M-token window and tiered pricing. How the billing works and when it beats renting the open weights.
Read
ModelsQwen3-Max Guide: Alibaba's Trillion-Parameter Closed Model
Qwen3-Max was Alibaba's largest model at launch: a closed flagship of over one trillion parameters from September 2025. What it is, and where it fits now.
ReadModelsQwen3-Max-Thinking: Reasoning With Tools It Picks Itself
Qwen3-Max-Thinking adds step-by-step reasoning, multi-round test-time scaling and self-selected search and code tools to Qwen3-Max. How to read its scores and when to pay.
ReadModelsQwen3-Next 80B A3B Instruct: Hybrid Attention, 3B Active
An 80B model that runs 3B parameters per token and swaps most attention layers for Gated DeltaNet. What that buys, where it trails the 235B, and how to serve it.
Read
ModelsQwen3-Next 80B A3B Thinking: Cheap Reasoning, Long Traces
A thinking-only open model with 3B active parameters that Qwen reports ahead of Gemini 2.5 Flash Thinking on maths and code. How to parse it, budget it and test it.
ReadModelsQwen3-VL 235B A22B Instruct: The Open Vision Flagship
Qwen's largest vision-language model reads documents, screens and hours of video without the usual drop in text quality. What it does well and what it costs to run.
ReadModelsQwen3-VL 235B A22B Thinking: When Images Need Reasoning
The reasoning edition of Qwen's largest vision model leads on visual maths but costs about twice as much per token as Instruct. Which image tasks justify that.
ReadModelsQwen3-VL 30B A3B Instruct: A Vision Model You Can Host
A 30B mixture-of-experts vision model with 3B active parameters, Apache-2.0 weights and an FP8 build. Where it is good enough, and where the 235B earns its cost.
Read
ModelsQwen3-VL 30B A3B Thinking: Visual Reasoning on a Budget
The small Qwen3-VL reasoning model gains more on code from thinking than the 235B does, at 3B active. Where it fits between the 30B Instruct and the 235B Thinking.
Read
ModelsQwen3-VL-32B Instruct: The Dense Vision Model Worth Keeping
Qwen3-VL-32B is the largest dense model in the Qwen3-VL line. What it does well, what Qwen3.5 changed, and when its low hosted price still makes it the pick.
ReadModelsQwen3-VL-8B Instruct: Small Enough to Run, Capable Enough to Use
Qwen3-VL-8B Instruct is an Apache-2.0 vision model you can self-host on modest hardware. What it handles, where it falls short, and how it compares with its siblings.
ReadShowing 265–276 of 446 articles