So Alibaba has just released a coding plan with all the best open source models You’ve access to Qwen 3.5-Plus, Kimi K2.5, MiniMax M2.5, GLM-5, etc. And you can plug it to: - Claude Code (as I did) - OpenClaw - Codex - Cline… While being waaay cheaper than... the paid plan of Codex, Claude & Cie. More belowshow more

It’s starting at $3 the first month (yes) and then it’s still at least 50% cheaper than the plans of the other big labs. And given how good the open source models are now, definitely worth it. Link:

the real value here is model diversity under one API key. when Qwen 3.5-Plus is better at certain tasks and Kimi K2.5 handles long context better, you can route different workloads to different models without managing 5 separate accounts. $3 first month is basically free to find out which OSS model fits your workflow best before committing

Yep, that's the good thing about it. You're not limited to a single model. And actually you even have access to image generation ones in the same plan.

So $3 for the first month and you get 18,000 API requests built in on Qwen 3.5-Plus, Kimi K2.5, MiniMax M2.5, GLM-5?

@LuciusFoxGME Yes, exactly!

I wonder whether it will include DeepSeek v4, if/when it drops...rumoured for tomorrow?? That would make an already superb value plan incredible!

And it is slow.

Not really

I've been there - paying extra for advanced models, only to realize most of the time I used the free ones available. It's great to see Alibaba making these resources more accessible.

Usually these ultra cheap API tiers come with some backend limitations I'd rather not hit at 2 AM

This one is backed by Alibaba Cloud which is just absolutely massive

Thanks Paul! Alibaba Cloud being massive is reassuring 👍 Will try it now

Love this. Cheap diverse models plus a good UX layer is the new IDE.

Democratizing access to high-compute models via open-source pipelines is the ultimate disruptor for proprietary gatekeeping. The cost-to-performance ratio here is a strategic goldmine. 🛠️💻

plugging Qwen 3.5-Plus into Claude Code for free is wild. open source eating the coding agent space

You cannot plug it into openclaw without braking the tos

Cheap is great, but model quality and tooling still decide everything. Have you stress-tested it on real projects yet?

Kimi GLM and MiniMax last models are very very good at tool calling. Zero issue.

Wow, this is a game-changer! Access to such a powerful lineup at a fraction of the cost opens up so many possibilities. Thanks for sharing can’t wait to experiment with these tools!

Interoperability is huge. Pricing pressure on closed plans just got real.

Doesn't work

? What kind of problem do you have?

Billing will not take my cards. Says for "new users only" and technically I have had an account with them for about 6 months But I've never been a paying customer Made a new account, still doesn't work But I mean, we all got problems, right?

If you’re posting AI content, make it feel like a conversation—not a brochure—because people engage with clarity, personality, and a little

$3 for first month seems attractive, however what about the rate for the next month? they even does not release anything about that in the pricing page. That is typical Ali, provide you with bait at first, then eat you alive

Open source is quietly eating margin.

I'm here for it honestly 👀

Same. 🙂 It’s forcing everyone to actually differentiate instead of relying on pricing power.

Cheaper is one thing but is it useful?

Well if you're using a coding agent or openclaw it is!

Have you benchmarked these inside Claude Code or OpenClaw for complex refactors or multi-file reasoning?

The gap between open and closed models is shrinking fast

Almost non-existent depending on the use cases

Cheap models don't matter if the cognitive cost is high. Switching between Claude and Qwen trains different patterns. The real question is whether you're building muscle memory or just saving money per token.

love your sharing! Really want to collaborate with you!

@itsPaulAi, this opens up some interesting possibilities for LLM inference.

How is the response time on it and tps vs getting the actual plans from minimax or kimi themselves?

I didn't compare the tps between the providers tbh. Seems really decent so far.

Can you use it with VS Code and Copilot?

How does it compare with sonnet 4.6 ? Or Opus 4.6

GLM-5 (which is available in the coding plan) is really close to Opus in my experience. Kimi is solid as well, a bit different. And MiniMax between Sonnet and Opus I'd say.

Wow. I need to try them out then. Interesting

Tools like this are making solo creators incredibly powerful.

Found the easiest way to learn @Alibaba_Qwen with @lmstudio in 10 mins:

plugging into Claude Code sounds like a great way to get started with those open source models

It is actually. Perfect to test the too calling as well.

Alibaba just made experimenting with Claude, Codex & friends way cheaper and way easier.

@AmolParikh40001 Exactly. Probably one of the plan that I will use the most tbh.

Very cool :)

Alibaba claw

Thanks

Welcome!

Not ready to feed my data to China.