Microsoft Favors Anthropic over OpenAI for Visual Studio Code
40 comments
·September 16, 2025kerpal
dmurray
Some people might be surprised that MS would pick the product with the best technological fit rather than the one they already have a deep business and financial relationship with.
Surely Microsoft's expertise these days is in cross-selling passable but not best-in-class products to enterprises who already pay for Microsoft products.
It says something about how they view the AI coding market, or perhaps the level of the gap between Anthropic and OpenAI here, that they've gone the other way.
pnathan
I've been happy with Anthropic models. I also have been using the Google models more, with decent results. The Copilot/OpenAI models don't seem to be as good as a rule of thumb, can't explain exactly why.
Overall, I think Google has a better breadth of knowledge encoded, but Anthropic gets work done better.
_fat_santa
This has been largely my experience as well. Claude does way better with coding while ChatGPT does better with general questions.
bobbylarrybobby
The new gpt-codex-* models are giving Claude Code a serious run for its money IMO. If OpenAI can figure out the Codex CLI UI (better permissions, more back and forth before executing) then I think they will have the better agentic coder.
m_mueller
GPT-5 is pretty decent nowadays, but Claude 4 Sonnet is superior in most cases. GPT beats it in cost and usable context window when something quite complex comes up to plan top-down.
CharlieIsAHero
What do you mean by usable context window? Sonnet 4 is 968k and gpt5 is 368k. Are you saying the context window on sonnet is useless?
CuriouslyC
Sonnet long context performance sucks. https://fiction.live/stories/Fiction-liveBench-Feb-21-2025/o...
I can confirm Sonnet is good for vibe coding but makes an absolute mess of large and complex codebases, while GPT5 tends to be pretty respectful.
m_mueller
I never implied it's useless. I don't have scientific data to back this up either, this is just my personal "feeling" from a couple hundred hours I've spent working with these models this year: GPT-5 seems a bit better at top-down architectural work, while Sonnet is better at the detail coding level. In terms of usable context window, again from personal experience so far, to me GPT-5 has somewhat of an edge.
boredtofears
What I find interesting is how much opinions vary on this. Open a different thread and people will seem to have consensus on GPT or Gemini being superior.
Even the bench marks don’t seem all that helpful.
TuxSH
Well, last I checked Claude's webchat UI doesn't have LaTeX rendering for output messages which is extremely annoying.
On the other hand, I wish ChatGPT had GitHub integration in Projects, not just in Codex.
kissgyorgy
I think it depends on the domain. For example, GPT-5 is better for frontend, React code, but struggles with niche things like Nix. Claude's UI designs are not as pretty as GPT-5's.
glimshe
Anthropic doesn't allow me to use my phone number across my personal and business logins. I simply can't use Claude where I need it, even if I'm willing to pay. I don't understand why they add so much friction when everyone else just allows me to do work.
electric_muse
This whole “real phone number is your access code to every service” trend is really frustrating.
I had the same experience recently with: - Ticketmaster - Docusign - Vercel
Probably a handful more I forgot.
I believe the main reason is because it prevents fraud.
But I see a deeper motive that phone numbers are more friction to change and therefore our “real” numbers become hard-to-change identity codes that can easily be used to pull tons of info about you.
You give them that number and they immediately can look up your name, addresses, age, and tons of other mined info that was connected to you. Probably credit score, household income, etc.
Phone numbers have tons of “metadata” you provide without really knowing it. Like how the Exif data in a photo may reveal a lot about your location and device.
derekdahmer
As someone who implemented phone verification at a company I worked for, it’s 100% for preventing spam signups intending to abuse free tiers. API companies can get huge volumes of fake signups from “multiplexers” who get around free tier limits by spreading their requests across multiple accounts.
AlexandrB
This makes sense for free tiers of products, but if you provide CC info for a paid tier, you shouldn't also have to provide a phone number. One or the other.
anonym29
Mandatory phone number registration does not and never has prevented fraud.
Plenty of free VOIP services exist, including SMS reception.
Even when the free service providers are manually blocklisted, one-time validations can be defeated with private numbers on real networks / providers for under a dollar per validation, and repeated ongoing validations can be performed with rented private numbers on real networks / providers for under ten dollars per month.
The rent-an-SMS services that enable this are accessible through a web interface that allows connections from tor, vpns, etc - there is no guarantee that the telecom provider's location records of the IMEI tied to that phone number is anywhere close to the end user's real geographic location, so this isn't even helpful for law enforcement purposes where they can subpoena telecom provider records.
This "phone number required" practice exists for one primary reason: for businesses to track non-fraudulent users, data mine their non-fraudulent users, and monetize the correlated personal information of non-fraudulent users without true informed consent (almost nobody reads ToS's, but many would object to these invasive practices if given a dialogue box that let them accept or decline the privacy infringements but still allowed the user to use the business' service either way).
Sometimes, they are also used for a secondary reason: to allow the business to cheap out on developer costs by cutting corners on proper, secure MFA validation. No need to implement modern, secure passkeys or RFC-compliant TOTP MFA, FIDO2, U2F when you can just put your users in harm's way by pretending that SMS is a secure channel, rather than easily compromised by even common criminals with SS7 attacks, which are not relegated to nation-state actors like they once were.
giancarlostoro
Sounds like you want a Google Voice Number or similar service, but now you're spending money for someone else's awful software, and in some cases, some places will flag your number if its google voice and outright refuse to let you in.
rs186
...Like Claude. They don't allow you to use Google voice numbers for verification.
giancarlostoro
I want a "burner" number, but I'm not sure what the best option is, do I buy a crappy phone at Walmart and use that number? What's the bare bottom of the barrel cost for a phone with no mobile data, only SMS?
dathinab
wait what do they need a phone number for???
catlover76
[dead]
dawnerd
Auto feels like a way for them to slightly push people towards paid models more. If they really favored Anthropic, claude would be the included free model, right?
paxys
Claude was the gold standard for coding but I have had a lot of success with GPT-5. Nowadays I pretty much always default to GPT-5.
jasonmarks_
Dueling r&d marketing budgets?
bwat49
yeah I've been getting better results with codex (gpt5) vs claude
piker
In some ways it makes sense to pave the way for Claude to protect the brand of VS Code. On the other hand, it’s a bit of a head-scratcher since it seems like VS code was built as a loss-leader to sell Microsoft cloud products. Perhaps enterprise ChatGPT, co-pilot and GitHub can make up the difference even if the community tier favors Claude.
Edit: maybe Cursor forced this and Microsoft is taking its choice to open license VS code on the chin. Will be interesting to see the strategy with Visual Studio going forward.
verdverm
Anthropic didn't make the cut in our evaluation (data usage concerns). They have also been the shadiest of the companies
They lost me when they expired my money and then tried to take more without asking
bartalama
Claude Sonnet 4 is the best for generating code for me so far, albeit needing some investment in instruction files and prompt files when using GitHub Copilot.
daft_pink
This needs that archive link that bypasses the paywall. I had to read it on my Apple News+ subscription to avoid the paywall.
bgarbiak
daft_pink
Thanks. Do I just enter the url into the top in this website to generate myself for future items?
cwkirchner
You can enter URLs at the top of archive.is (or archive.ph) to see the archived version. Alternatively, there's a bookmarklet that you can use for archive.ph.
gigel82
I don't think it's Microsoft that favors it. It's likely customers. Claude wipes the floor with all the GPTs in GitHub Copilot (in my experience).
Claude/Anthropic is more focused on productivity (Coding, Spreadsheets, Reports). ChatGPT seems more focused on general-purpose LLM (Research, Cooking, Writing, Image Generation).
Makes sense that MS would partner with Anthropic since their tool-use for productivity (Claude Code) seems superior. I personally rarely code with ChatGPT, almost strictly Claude.