Qwen3-Coder-Next-GGUF, it is surprisingly good
March 23, 2026 at 10:21 PM
Qwen3-Coder-Next-GGUF, it is surprisingly good
Just put Qwen3-Coder on your machine, problem solved
When I see LLMs pitched as a guaranteed productivity tool for software engineering, the is the counterweight I keep trying to get people to seriously engage with and not handwave away is the future cost, and “oh everything tech gets cheaper” isn’t an answer. Services don’t and LLMs are a service.
November 30, 2025 at 5:05 PM
Just put Qwen3-Coder on your machine, problem solved
In writing up today's release of Qwen3-Coder-30B-A3B-Instruct - the 6th model released by Qwen this July! - I ended up putting together a tutorial on using LM Studio and Open WebUI and LLM and mlx-lm to run the model on a 32GB or 64GB Mac simonwillison.net/2025/Jul/31/...
Trying out Qwen3 Coder Flash using LM Studio and Open WebUI and LLM
Qwen just released their sixth model(!) for this July called Qwen3-Coder-30B-A3B-Instruct—listed as Qwen3-Coder-Flash in their chat.qwen.ai interface. It’s 30.5B total parameters with 3.3B active at a...
simonwillison.net
July 31, 2025 at 7:58 PM
In writing up today's release of Qwen3-Coder-30B-A3B-Instruct - the 6th model released by Qwen this July! - I ended up putting together a tutorial on using LM Studio and Open WebUI and LLM and mlx-lm to run the model on a 32GB or 64GB Mac simonwillison.net/2025/Jul/31/...
Hey I can run this thing without data centers, and wasting any water.
February 7, 2026 at 5:18 PM
Hey I can run this thing without data centers, and wasting any water.
here it is:
* benchmarks: tough competition with Sonnet-4
* 256K context, expandable to 1M with YaRN
there’s also a CLI forked from gemini-cli
qwenlm.github.io/blog/qwen3-c...
* benchmarks: tough competition with Sonnet-4
* 256K context, expandable to 1M with YaRN
there’s also a CLI forked from gemini-cli
qwenlm.github.io/blog/qwen3-c...
July 22, 2025 at 9:44 PM
here it is:
* benchmarks: tough competition with Sonnet-4
* 256K context, expandable to 1M with YaRN
there’s also a CLI forked from gemini-cli
qwenlm.github.io/blog/qwen3-c...
* benchmarks: tough competition with Sonnet-4
* 256K context, expandable to 1M with YaRN
there’s also a CLI forked from gemini-cli
qwenlm.github.io/blog/qwen3-c...
Qwen does some weird layer-looped upcycling to make some models, especially Qwen3-Coder and Qwen3-Max
November 5, 2025 at 10:00 PM
Qwen does some weird layer-looped upcycling to make some models, especially Qwen3-Coder and Qwen3-Max
Cerebras is deprecating Qwen3-Coder-480b and Qwen3-235B-A22B
but they’re bringing in GLM-4.6
but they’re bringing in GLM-4.6
October 23, 2025 at 10:52 AM
Cerebras is deprecating Qwen3-Coder-480b and Qwen3-235B-A22B
but they’re bringing in GLM-4.6
but they’re bringing in GLM-4.6
Qwen3-Coder-Next ist ein neues KI-Modell von Alibaba, das speziell für Programmier-Agenten und lokale Entwicklungs-Workflows entwickelt wurde.
In Tests erzielt Qwen3-Coder-Next vergleichbare oder bessere Leistung als deutlich größere Open-Source-Modelle.
www.marktechpost.com/2026/02/03/q...
In Tests erzielt Qwen3-Coder-Next vergleichbare oder bessere Leistung als deutlich größere Open-Source-Modelle.
www.marktechpost.com/2026/02/03/q...
Qwen Team Releases Qwen3-Coder-Next: An Open-Weight Language Model Designed Specifically for Coding Agents and Local Development
Qwen Team Releases Qwen3-Coder-Next: An Open-Weight Language Model Designed Specifically for Coding Agents and Local Development
www.marktechpost.com
February 5, 2026 at 4:57 PM
Qwen3-Coder-Next ist ein neues KI-Modell von Alibaba, das speziell für Programmier-Agenten und lokale Entwicklungs-Workflows entwickelt wurde.
In Tests erzielt Qwen3-Coder-Next vergleichbare oder bessere Leistung als deutlich größere Open-Source-Modelle.
www.marktechpost.com/2026/02/03/q...
In Tests erzielt Qwen3-Coder-Next vergleichbare oder bessere Leistung als deutlich größere Open-Source-Modelle.
www.marktechpost.com/2026/02/03/q...
Qwen released their updated "thinking" model today. It thinks really hard! Took 166 seconds to think through the details of drawing me a pelican on a bicycle. The finished drawing wasn't great but the thoughts behind it were fun to see.
simonwillison.net/2025/Jul/25/...
simonwillison.net/2025/Jul/25/...
Qwen3-235B-A22B-Thinking-2507
The third Qwen model release week, following Qwen3-235B-A22B-Instruct-2507 on Monday 21st and Qwen3-Coder-480B-A35B-Instruct on Tuesday 22nd. Those two were both non-reasoning models - a change from t...
simonwillison.net
July 25, 2025 at 10:53 PM
Qwen released their updated "thinking" model today. It thinks really hard! Took 166 seconds to think through the details of drawing me a pelican on a bicycle. The finished drawing wasn't great but the thoughts behind it were fun to see.
simonwillison.net/2025/Jul/25/...
simonwillison.net/2025/Jul/25/...
Qwen3-Coder-Flash: Qwen3-Coder-30B-A3B-Instruct
- Native 256K context (supports up to 1M tokens with YaRN)
- Optimized for platforms like Qwen Code, Cline, Roo Code, Kilo Code, etc.
💬 Chat: chat.qwen.ai
🤗 Model: hf.co/Qwen/Qwen3-C...
🔧 Repo: github.com/QwenLM/qwen-...
- Native 256K context (supports up to 1M tokens with YaRN)
- Optimized for platforms like Qwen Code, Cline, Roo Code, Kilo Code, etc.
💬 Chat: chat.qwen.ai
🤗 Model: hf.co/Qwen/Qwen3-C...
🔧 Repo: github.com/QwenLM/qwen-...
July 31, 2025 at 2:50 PM
Qwen3-Coder-Flash: Qwen3-Coder-30B-A3B-Instruct
- Native 256K context (supports up to 1M tokens with YaRN)
- Optimized for platforms like Qwen Code, Cline, Roo Code, Kilo Code, etc.
💬 Chat: chat.qwen.ai
🤗 Model: hf.co/Qwen/Qwen3-C...
🔧 Repo: github.com/QwenLM/qwen-...
- Native 256K context (supports up to 1M tokens with YaRN)
- Optimized for platforms like Qwen Code, Cline, Roo Code, Kilo Code, etc.
💬 Chat: chat.qwen.ai
🤗 Model: hf.co/Qwen/Qwen3-C...
🔧 Repo: github.com/QwenLM/qwen-...
Qwen3 Coder is WILD
December 16, 2025 at 3:00 AM
Qwen3 Coder is WILD
The post they are replying to was written on an MacBook Pro I upgraded to 64 GB of RAM specifically to be able to run 70B models
You can run Qwen3-Coder-30B on 32GB or RAM, and it's surprisingly capable at most coding tasks, even
You can run Qwen3-Coder-30B on 32GB or RAM, and it's surprisingly capable at most coding tasks, even
December 24, 2025 at 3:52 AM
The post they are replying to was written on an MacBook Pro I upgraded to 64 GB of RAM specifically to be able to run 70B models
You can run Qwen3-Coder-30B on 32GB or RAM, and it's surprisingly capable at most coding tasks, even
You can run Qwen3-Coder-30B on 32GB or RAM, and it's surprisingly capable at most coding tasks, even
軽量で高性能なコーディングエージェント「Qwen3-Coder-Next」が登場
https://gigazine.net/news/20260204-alibaba-qwen3-coder-next/
https://gigazine.net/news/20260204-alibaba-qwen3-coder-next/
軽量で高性能なコーディングエージェント「Qwen3-Coder-Next」が登場
アリババが、コーディングエージェントとローカル開発用に構築されたオープンウェイト言語モデル「Qwen3-Coder-Next」を発表しました。総パラメーター数は80B(800億)で、推論コストを大幅に低減しつつ強力なコーディング能力とエージェント能力を獲得しています。
gigazine.net
February 4, 2026 at 2:37 AM
軽量で高性能なコーディングエージェント「Qwen3-Coder-Next」が登場
https://gigazine.net/news/20260204-alibaba-qwen3-coder-next/
https://gigazine.net/news/20260204-alibaba-qwen3-coder-next/
the 480B was exclusively released as an instruction-tuned "Qwen3-Coder" and the 1T model (proprietary, "Qwen3-Max") is so deeply fried that it does not know the difference between "kaomoji" and "emoji"
November 5, 2025 at 10:20 PM
the 480B was exclusively released as an instruction-tuned "Qwen3-Coder" and the 1T model (proprietary, "Qwen3-Max") is so deeply fried that it does not know the difference between "kaomoji" and "emoji"
hmm my entire python codebase only takes up like 30% of qwen3 coder next's context
February 9, 2026 at 7:00 PM
hmm my entire python codebase only takes up like 30% of qwen3 coder next's context
Jupyter Agent Dataset
Built from 7 TB of real Kaggle datasets + 20k notebooks, creating real code exec traces using Qwen3-Coder and E2B.
huggingface.co/datasets/dat...
Built from 7 TB of real Kaggle datasets + 20k notebooks, creating real code exec traces using Qwen3-Coder and E2B.
huggingface.co/datasets/dat...
September 3, 2025 at 12:40 AM
Jupyter Agent Dataset
Built from 7 TB of real Kaggle datasets + 20k notebooks, creating real code exec traces using Qwen3-Coder and E2B.
huggingface.co/datasets/dat...
Built from 7 TB of real Kaggle datasets + 20k notebooks, creating real code exec traces using Qwen3-Coder and E2B.
huggingface.co/datasets/dat...
I've got dual qwen3-coder-30B agents on my AI Max+ 395 now. This is the promised land.
killed this because
1. I couldn't get streaming tool calling to work perfectly
2. I was able to fix jinja template for qwen3-coder on branch of llama.cpp I am using for hosting glm-4.5-air
1. I couldn't get streaming tool calling to work perfectly
2. I was able to fix jinja template for qwen3-coder on branch of llama.cpp I am using for hosting glm-4.5-air
I am writing my own llama-cpp-python server because the one they wrote tries to load the whole model into cpu memory, which fails for my case. The local version works fine, so I am buliding a single-threaded llama.cpp openai compliant server smdh.
November 1, 2025 at 4:32 PM
I've got dual qwen3-coder-30B agents on my AI Max+ 395 now. This is the promised land.
Qwen3-coder seems to be great 👀
July 22, 2025 at 10:40 PM
Qwen3-coder seems to be great 👀
In writing up today's release of Qwen3-Coder-30B-A3B-Instruct - the 6th model released by Qwen this July! - I ended up putting together a tutorial on using LM Studio and Open WebUI and LLM and mlx-lm to run the model on a 32GB or 64GB Mac https://simonwillison.net/2025/Jul/31/qwen3-coder-flash/
Trying out Qwen3 Coder Flash using LM Studio and Open WebUI and LLM
Qwen just released their sixth model(!) for this July called Qwen3-Coder-30B-A3B-Instruct—listed as Qwen3-Coder-Flash in their chat.qwen.ai interface. It’s 30.5B total parameters with 3.3B active at any one time. This means …
simonwillison.net
July 31, 2025 at 8:01 PM
In writing up today's release of Qwen3-Coder-30B-A3B-Instruct - the 6th model released by Qwen this July! - I ended up putting together a tutorial on using LM Studio and Open WebUI and LLM and mlx-lm to run the model on a 32GB or 64GB Mac https://simonwillison.net/2025/Jul/31/qwen3-coder-flash/
The new Qwen3-Coder is out. It’s currently the best and mos powerful open-source model available, with 480 billion parameters (35 billion active), support for 358 programming languages, and long-context capabilities. Most importantly, it’s completely free and open source.
July 23, 2025 at 6:09 AM
The new Qwen3-Coder is out. It’s currently the best and mos powerful open-source model available, with 480 billion parameters (35 billion active), support for 358 programming languages, and long-context capabilities. Most importantly, it’s completely free and open source.
Alibaba's Qwen3-Coder-Next, an open-weight LM built for coding agents & local development.
🤖 Scaling agentic training: 800K verifiable tasks + executable envs
📈 Efficiency–Performance Tradeoff: achieves strong results on SWE-Bench Pro with 80B total params and 3B active
🤖 Scaling agentic training: 800K verifiable tasks + executable envs
📈 Efficiency–Performance Tradeoff: achieves strong results on SWE-Bench Pro with 80B total params and 3B active
February 3, 2026 at 5:03 PM
Alibaba's Qwen3-Coder-Next, an open-weight LM built for coding agents & local development.
🤖 Scaling agentic training: 800K verifiable tasks + executable envs
📈 Efficiency–Performance Tradeoff: achieves strong results on SWE-Bench Pro with 80B total params and 3B active
🤖 Scaling agentic training: 800K verifiable tasks + executable envs
📈 Efficiency–Performance Tradeoff: achieves strong results on SWE-Bench Pro with 80B total params and 3B active
I’m _very_ curious about the small positive skew Qwen3-Coder-Next has. Exercising extreme restraint not digging into it, staying focused on the current goal.
February 13, 2026 at 1:22 PM
I’m _very_ curious about the small positive skew Qwen3-Coder-Next has. Exercising extreme restraint not digging into it, staying focused on the current goal.
docs.unsloth.ai/basics/qwen3...
after all, why *shouldn't* I buy 256GB of DDR4-3200 and put it in my server? it's only... 700 USD... okay, maybe not
after all, why *shouldn't* I buy 256GB of DDR4-3200 and put it in my server? it's only... 700 USD... okay, maybe not
Qwen3-Coder: How to Run Locally | Unsloth Documentation
Run Qwen3-Coder-480B-A35B locally with Unsloth Dynamic quants.
docs.unsloth.ai
July 23, 2025 at 12:55 PM
docs.unsloth.ai/basics/qwen3...
after all, why *shouldn't* I buy 256GB of DDR4-3200 and put it in my server? it's only... 700 USD... okay, maybe not
after all, why *shouldn't* I buy 256GB of DDR4-3200 and put it in my server? it's only... 700 USD... okay, maybe not
Looks like at least one of my machines might have to become dedicated to running Qwen3-Coder-Next. Wow.
Might be time to finally give a fully local Opencode a shot.
Might be time to finally give a fully local Opencode a shot.
February 4, 2026 at 2:36 PM
Looks like at least one of my machines might have to become dedicated to running Qwen3-Coder-Next. Wow.
Might be time to finally give a fully local Opencode a shot.
Might be time to finally give a fully local Opencode a shot.