Please add the new Qwen Instruct model on agent and MAX mode. It apparently far surpasses Kimi-K2, and Opus4 a little on some benchmarks. Cheap price and large enough context window:
262,144 context
$0.15/M input tokens
$0.85/M output tokens
39 Likes
Please also add Minimax-m1 as well - it’s been around for a short while now and is great too:
1,000,000 context
Starting at $0.30/M input tokens
Starting at $1.65/M output tokens
3 Likes
+++ hopefully we get some traction on this post
1 Like
you can use openrouter for this
This is ridiculous. Cursor should come with top-tier models, not have to buy them through third parties. With this approach, you could argue that you’d need to purchase separate products for everything else – so why pay for Cursor at all?
8 Likes
I believe I asked to add this model it will not work Deep example is an excellent model The cost could very well leave it unlimited for us since he pays directly to firework and even so he preferred to limit even the EPSEAK itself
New LLM: Coming out
Cursor users after 0.00001 second:
5 Likes
yeah but shouldn’t have to. Plus using models via openrouter doesn’t necessarily give you access to agent and/or MAX mode. These models are super efficient and cheap, on par if not better than the best proprietary models so why not serve directly to users ? We’re already paying Pro and PRo+ subscriptions, so might as well keep us happy 
1 Like
time is of the essence brethren
We need these !
3 Likes
Qwen has been on fire these past 2 days: https://x.com/Alibaba_Qwen/status/1947766835023335516
So yeah, please add this one too to Cursor
Pricing is out on openrouter:
1$/M tokens
9 Likes
Just used Qwen3-coder (1M context) with agent mode on a large task (about 30 files) to find a bug, via openrouter. It works perfectly, uses the tools, etc.
Also tested the Instruct model, it too works but is a bit slower.
4 Likes
The reality is that these open models all suc.ks, even Gemini is inferior to Claude. just waste of time. I asked a simple task to kimi2 and it did it completely wrong. Asked the same to Claude and it nailed it, like perfectly. For now, claude is the unique production grade model for programmers. this is my opinion.
These benchmarks are far way from reality and agent capabilities. So, guys, don’t rush for these new models.
6 Likes
Hi Cursor team,
I’d love to see the Qwen 3-Coder model integrated into AUTO mode for users on the PRO+ or ULTRA plans. It offers exceptional performance and efficiency—ideal for code-heavy workflows.
Given the recent inclusion of other models like Qwen3-235B-A22B and Minimax-M1, adding Qwen 3-Coder would significantly enhance the experience for developers who rely on automatic agent switching to match tasks to the best model.
Its strong capabilities in code generation, reasonable pricing, and large context window make it a perfect candidate.
Please consider supporting Qwen 3-Coder in AUTO mode for premium plan users.
Thanks!
Or better instead of Cursor-small
1 Like
Thanks for the suggestion, we will consider this!
7 Likes
I have tested in real production
this working better, creative
I try to redesign existing eLearning
Claude4 sonnet got bad result but Qwen3 Coder give me better result with same prompt
this can be an alternative for Claude AI Agent
4 Likes
I’m genuinely hoping for changes, because at this point I don’t see any point in buying even the ULTRA plan. I spoke with someone who purchased it, and they said it’s much more cost-effective to use Cloud Code, since even with ULTRA the usage limits are reached very quickly, and the AUTO mode is nearly impossible to use.
We really need unlimited access to this model. It’s publicly available, so you could deploy it on your own servers—at least for PRO+ or ULTRA users. If you were to offer that, I would not only continue paying for the subscription, but many others who think the same way would do the same.
1 Like
I think this is a nice suggestion but to add a bit
I think models like Kimi K2 and Qwen3 should be unlimited for Pro+ and Ultra Plans
3 Likes
I meant exactly that, because nowadays many people don’t see the point in renewing their subscription.
1 Like