Run Qwen3 Coder on W&B Inference
Get started with the latest Qwen3 models on our free tier.
Created on July 24|Last edited on July 25
Comment
W&B Inference, powered by CoreWeave, now provides API and playground access to the Qwen3 Coder 480B A35B and Qwen3 235B A22B-2507 models with one of the lowest costs per 1M tokens in the industry. These two Qwen3 models are Alibaba’s latest open-source mixture-of-experts models, optimized respectively for coding and general language tasks.
Instantly access these new Qwen3 models fully hosted on CoreWeave's AI Cloud platform using just the OpenAI API. Quickly evaluate, monitor, and iterate on your agentic AI applications by leveraging integrated W&B Weave tracing available through W&B Inference.
Qwen3 Coder beats Kimi K2 and on par with Claude Sonnet 4
Qwen3 Coder 480B A35B has 480B parameters (with 35B active parameters) and supports a native 256K context length, extendable up to 1M through extrapolation methods. It achieves state-of-the-art results among open models for agentic coding, browser interaction, and tool use—performance comparable to Claude Sonnet 4 and superior to Kimi K2.
Qwen3 235B A22B Instruct-2507, a 235B parameter model, excels in general-purpose capabilities such as instruction-following and logical reasoning, surpassing Kimi K2 on benchmarks including GPQA and AIME25. It’s an update to the Qwen3 model released a few months ago and beats Kimi K2 released just 9 days ago.
Serverless access on W&B Inference powered by CoreWeave
Skip the hassle of signing up for one more model hosting provider or deploying the model yourself. Your Weights & Biases account gives you instant access to the new Qwen models and other top open-source foundation models, fully hosted on powerful CoreWeave infrastructure. What's more, our pricing is one of the industry’s best so you can evaluate and use the models in your applications and workflows cost-effectively. Simply sign in to Weights & Biases, select one of the Qwen3 models from the menu, and start running inference for free in seconds.
Easiest way to try the models instantly is via the W&B Weave Playground—no model endpoints, no access keys, and zero configuration needed.
If you want to run it from your code, navigate to the model card and copy and paste the starter code we’ve provided.
Tool calling support
Tool calling (or function calling) is an essential feature for maximizing the effectiveness of agentic models. That's why we've added support for tool calling when accessing models through W&B Inference. This capability is especially valuable when working with coding-focused models like Qwen3 Coder. Simply provide a list of tools when calling the OpenAI API, and the model will automatically select the appropriate tool from your list. To help you get started, we've included an example of tool calling in our sample Colab notebook.
Tool calling support is currently available for Qwen3 Coder and all other models on W&B Inference except Kimi K2. Support for tool calling with Kimi K2 is coming soon.
Getting started
Try Qwen3 Coder and Qwen3 Instruct right now through the W&B Weave Playground. Every Weights & Biases plan includes a free tier of W&B Inference, so you can dive straight in without additional upfront costs. To learn more, see the W&B Inference documentation and the W&B Inference pricing page.
Add a comment
Iterate on AI agents and models faster. Try Weights & Biases today.