Ollama 0.1.5: Qwen3-Coder-Next and the Joys of Running Models Locally

Ollama has been updated to version 0.15.5, adding support for new models. The standout is Qwen3-Coder-Next. Built for code generation, it promises to be a powerful tool for developers, but as is often the case, high performance comes with high resource demands.

Qwen3-Coder-Next is a model with 80 billion parameters, most of which were trained on code. That lets it deliver impressive results on programming tasks. Running it locally, however, takes serious hardware. Ollama offers only quantized versions of the model (q4_K_M at 52 GB and q8_0 at 85 GB), which already hints at how resource-hungry it is.

For comfortable work with Qwen3-Coder-Next you will need no less than 80 GB of VRAM if you want high inference speed. The model can also run on a CPU with 128 GB of DDR5 RAM or more, but speed will be significantly lower. That makes Qwen3-Coder-Next far from the most accessible option for an average user, which, as the author of the original article notes, often provokes a negative reaction.

Interestingly, the author offers free access to their own server, which already has 10 different models loaded, including Qwen3-Coder-Next. It is a good opportunity for anyone who cannot run the model locally but wants to try it out. Keep in mind, though, that the access is not permanent and will be closed once the setup of a RAG (Retrieval-Augmented Generation) system is finished. The author also warns that they will cut off users who try to overload the server.

Overall, the arrival of Qwen3-Coder-Next in Ollama is an important step forward for local LLMs. It lets developers use a powerful code-generation model without relying on cloud services. Still, the high resource requirements remain a serious obstacle to wider adoption of such models. The author's offer of free server access is a good opportunity for those who want to test Qwen3-Coder-Next but lack the necessary hardware. It also highlights a growing trend toward local, self-sufficient AI solutions that give users more control over their data and computing resources.

Ollama, then, keeps evolving and gives users access to cutting-edge models, but it is worth remembering that hardware requirements keep growing. Trying Qwen3-Coder-Next on the author's server is a valuable chance to gauge the potential of such models and decide whether you are ready to move to more powerful hardware.