Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

MAI-Code-1-Flash

microsoft.ai

Post reply on HN
Print view View original

231–240 of 297 posts

Re: MAI-Code-1-Flash

#231

Post by sfifs »

camelmel wrote: post #43

Huh, according to that model card this is a 137B total parameter model. Performance doesn't seem that good: - MAI-Code-1-Flash (137B-A5B) = 51% on SWE-bench pro - Qwen3.6-35B-A3B = 49.5% on SWE-bench pro ( https://huggingface.co/Qwen/Qwen3.6-35B-A3B ) They benchmark against Claude Haiku but Haiku is not good, it's worse than tiny open models you can run locally or via API at 10% the cost.

Qwen is definitely the model to beat as of Mid 2026. While I didn't benchmark with SWE as my use cases are OpenClaw [1]. I found both Qwen 3.6 35B A3B and more impressively Qwen 3.5 122B A10B starting to be competitive with closed flash models. The NVFP4 quant of the latter is what I'm running now on DGX.

[1] https://srinathh.medium.com/mid-size-local-models-are-now-co...

Re: MAI-Code-1-Flash

#232

Post by Hfuffzehn »

bel8 wrote: post #82

It's a start and I welcome competition but I don't think I ever used small cloud models like Haiku 4.5. They are cute but for serious coding they tend to waste your expensive time. And this certainly wont bring me back to GitHub Copilot which I cancelled yesterday. GitHub Copilot had competitive pricing until yesterday when they changed from per-request to one of the most expensive per-token quotas. Seriously, take a…

Agreed. Seems like this could have been a nice model if we would still be in the old GitHub Copilot free request/ premium multiplier mode. It could have been a good compromise to somehow reign in the costs for Microsoft.

But with Copilot now just being paying per-token prices I don't see how this is competitive with Chinese models.

It is probably telling you can't find the costs in the announcement. Because Input $0.75 Cached input $0.075 Output $4.50 might be competitive with Haiku, but nobody in their right mind uses Haiku and Anthropic has abandoned it chasing the tokenmaxers who aren't thinking about budgets.

So I guess they are aiming for corporate customers that are bound to Microsoft through compliance approval that will soon start seeing their budgets explode that have to find some corporate compromise.

Re: MAI-Code-1-Flash

#233

Post by claud_ia »

hmokiguess wrote: post #33

Does anyone actually uses these smaller models for coding? If so, how? I usually Opus everything. Is the play to plan/design/architect with a heavier model than delegate structured tasks to these smaller ones? Would appreciate to hear someone's opinion on having done and tested both paths.

[flagged]

Re: MAI-Code-1-Flash

#235

Post by ignoramous »

minraws wrote: post #62
giancarlostoro wrote: post #57

Earlier quoted context omitted.

The take away is that this model is a smaller model that competes with Haiku, I would hope they come out with a "Sonnet" competing model, then Opus. I have been wondering why Microsoft is kind of "sleeping" on offering models they themselves have made on Copilot, maybe it was part of their deal with OpenAI? Not sure.

They did release, MAI-Thinking-1 to compete with Sonnet. Totally not sure why that isn't at the top here.

Can't yet use MAI-Thinking-1? [0] And no indication of it being made available in GitHub Copilot, either.

[0] Not even here: https://playground.microsoft.ai/

Re: MAI-Code-1-Flash

#236

Post by kosolam »

davecitron wrote: post #211
camelmel wrote: post #43

Huh, according to that model card this is a 137B total parameter model. Performance doesn't seem that good: - MAI-Code-1-Flash (137B-A5B) = 51% on SWE-bench pro - Qwen3.6-35B-A3B = 49.5% on SWE-bench pro ( https://huggingface.co/Qwen/Qwen3.6-35B-A3B ) They benchmark against Claude Haiku but Haiku is not good, it's worse than tiny open models you can run locally or via API at 10% the cost.

Dave Citron here, from the MAI team. Thanks for the feedback, we're getting the model card updated to call out 5B active parameters (137B total). On benchmarks: in the same VS Code harness, MAI-Code-1-Flash scored 51.2% on SWE-bench Pro vs. Haiku's 35.2% which we see as a pretty big leap. But going forward, we'll include additional models in our benchmarks, including models like Qwen 3.6 and Gemma 4.

Hey Dave, I’d love to add your new model in the harness I’m going to opensource very soonish. Going to publish benchmarks on real world tasks.

Re: MAI-Code-1-Flash

#237

Post by ignoramous »

papascrubs wrote: post #168
GaryBluto wrote: post #101

Earlier quoted context omitted.

Almost exactly the same story here. I've also had little to no refusals from DeepSeek, with it's Chinese values meaning substantially less friction when it comes to things like reverse engineering, finding copyrighted files, working with dubiously-sourced source code, et cetera. I don't think I'd go back to Copilot even if they dropped prices by 90%.

Are you purchasing directly from DeepSeek? Any concerns as far as privacy or data protection?

> Any concerns as far as privacy or data protection?

We moved to OpenCode Go ($10/mo), so we could switch between DeepSeek v4, GLM 5.1, and Qwen 3.7 models run by providers in EU, US, & Singapore that OpenCode FAQ claims don't use retained data for training.

  What about data and privacy?

  The [OpenCode Go] plan is designed primarily for international users, with models hosted in the US, EU, and Singapore for stable global access. Our providers follow a zero-retention policy and do not use your data for model training.
I find their rather verbose privacy policy is not making far-reaching guarantees about any of this though: https://opencode.ai/legal/privacy-policy

Re: MAI-Code-1-Flash

#238

Post by ignoramous »

bel8 wrote: post #225
epolanski wrote: post #215

Earlier quoted context omitted.

> They are cute but for serious coding they tend to waste your expensive time. 90% of corporate job tasks are trivial enough that Haiku can handle them. Just this morning I have been implementing a reprint functionality in our warehouse management system, which needed to print again carrier labels and delivery notes for a specific order. It essentially had to do the same workflow of print, but instead of generating a…

Unfortunately there's no defending Haiku 4.5 at this point when cheaper and better options are available. TLDR: https://artificialanalysis.ai/models?models=gemini-3-5-flash... and: https://i.imgur.com/nTu3VCZ.png For starters I did experiment a heck lot with models since Github Copilot gave me access to OpenAI, Gemini and Anthropic models. So I probably experimented more than the average LLMer. When GitHub Copilot ha…

> DeepSeek Flash on high setting

In your experience, is max worse or you suggest it for less token use?

> MiMo 2.5 Pro on the same OpenCode Go

Xiaomi dropped dropped MiMo 2.5 rates by 70%+ [0] & now it is cost competitive with DeepSeek v4 Pro. I haven't used MiMo, but since you have, do you find it to be better than DeepSeek v4? If so, for what tasks? How do you decide when to use which, if you have an intuition for it? Thanks.

[0] https://news.ycombinator.com/item?id=48282814

Re: MAI-Code-1-Flash

#239

Post by Hfuffzehn »

roywiggins wrote: post #199
eli wrote: post #175

Earlier quoted context omitted.

Makes sense as part of a larger coding workflow, especially if it’s fast. Using a trillion parameter model to figure out how to call a targeted edit tool or generate a commit message is a waste. Also narrow tasks like “make the background darker” or “rename this function and update callers”

> “rename this function and update callers” I'm old enough to remember when IDEs could do this without needing a couple gigabytes of matrices to do it (LLMs are great for anything even slightly more complicated ofc)

The first time I was impressed by AI coding was when I pointed it at some switch case monster code and told it to replace it with a strategy pattern.

And it did just fine.

So no matter what you think about vibe coding, using AI for these slightly more complicated use cases is genuinely useful.

Re: MAI-Code-1-Flash

#240

Post by Hfuffzehn »

bgirard wrote: post #115
redrove wrote: post #27

Earlier quoted context omitted.

It’s about bang for buck. That high a score for 5B params is pretty good, nigh unbelievable a short while ago. It is my belief that smaller models will get better and better, and even cloud SOTA models will shrink. Yet another reason the current buildout will feel like the railroads.

> It’s about bang for buck. Hard to know when they don't give the price per token. Presumably it will be comparable to a low-mid range model in terms of price. But otherwise their 'Ideal Zone' is meaningless without factoring in the price per token. I don't how much tokens are being used, that's an implementation detail to me. I care about price / performance / latency.

https://docs.github.com/en/copilot/reference/copilot-billing...

Model Input Cached input Output MAI-Code-1-Flash $0.75 $0.075 $4.50

Post reply on HN