[HN Gopher] Zed's Pricing Has Changed: LLM Usage Is Now Token-Based
       ___________________________________________________________________
        
       Zed's Pricing Has Changed: LLM Usage Is Now Token-Based
        
       Author : meetpateltech
       Score  : 154 points
       Date   : 2025-09-24 16:13 UTC (6 hours ago)
        
 (HTM) web link (zed.dev)
 (TXT) w3m dump (zed.dev)
        
       | input_sh wrote:
       | Entirely predictable and what should've been done from the start
       | instead of this bait-and-switch mere months after introducing
       | agentic editing.
        
         | relativeadv wrote:
         | is this effectively what Cursor did as well? I seem to remember
         | some major pricing change of their in the past few months.
        
           | input_sh wrote:
           | In a way I would say they were even worse, instead of
           | outright saying "we've increased our prices", they "clarified
           | their pricing".
        
       | cactusplant7374 wrote:
       | How much are companies spending per developer on tokens? From
       | what I read it seems like it might be quite high at $1,000 or
       | more per day?
        
         | trenchpilgrim wrote:
         | No, not at all! At my org it's around $7000 a month for the
         | entire org - my personal usage is around $2-10 a day. Usually
         | less than the price of my caffeinated beverages.
        
       | binwang wrote:
       | Now I see little value in subscribing to Zed Pro compared to just
       | bringing my own API key. Am I missing something?
        
         | prasoon2211 wrote:
         | Presumably the tab based edit-prediction model + $5 of tokens
         | is worth the (new) $10 / mo price.
         | 
         | Though from everything I've read online, Zed's edit prediction
         | model is far, _far_ behind that of Cursor.
        
         | agrippanux wrote:
         | Their burn agent mode is pretty badass, but is super costly to
         | run.
         | 
         | I'm a big fan of Zed but tbf I'm just using Claude Code + Nvim
         | nowadays. Zed's problem with their Claude integration is that
         | it will never be as good as just using the latest from Claude
         | Code.
        
         | morgankrey wrote:
         | (I work at Zed) No, you aren't. We care about you using Zed the
         | editor, and we provide Zed Pro for folks who decide they'd like
         | to support Zed or our billing model works for them. But it's
         | simply an option, not our core business plan, and this pricing
         | is in place to make that option financially viable for us. As
         | long as we don't bear the cost, we don't feel the need (or the
         | right) to put ourselves in the revenue path with LLM spend.
        
           | jsheard wrote:
           | > [Zed Pro is] not our core business plan
           | 
           | What is the core business plan then?
        
             | morgankrey wrote:
             | https://zed.dev/blog/sequoia-backs-zed#introducing-
             | deltadb-o...
        
           | maxbond wrote:
           | Will you consider providing a feature to protect me from
           | accidentally using my Zed account after the $5 is exhausted
           | (or else a plan that only includes edit predictions)? I can't
           | justify to myself continuing my subscription if there's a
           | risk I will click the wrong button with identical text to the
           | right button, and get charged an additional 10% for it. I get
           | you need to be compensated for risk if you pay up front on my
           | behalf, but I don't need you to do that.
           | 
           | I understand that there's nothing you could do to protect me
           | if I make a prompt that ends up using >$5 of usage but after
           | that I would like Zed to reject anything except my personal
           | API keys.
        
             | morgankrey wrote:
             | Yep, you can set your spend limit to $0 and it will block
             | any spend beyond your $10 per month for the subscription
             | 
             | https://zed.dev/docs/ai/plans-and-usage#usage-spend-limits
        
               | maxbond wrote:
               | Excellent. Thanks.
        
           | drdaeman wrote:
           | I'm curious if there's any way to completely disable/remove
           | `zed.dev` provider from Zed, while keeping others available?
        
             | ibejoeb wrote:
             | If you sign out of zed, zed's providers don't work. I
             | believe you still see them in the AI panel, but it won't
             | operate.
        
       | bluehatbrit wrote:
       | Token based pricing generally makes a lot of sense for companies
       | like Zed, but it sure does suck for forecasting spend.
       | 
       | Usage pricing on something like aws is pretty easy to figure out.
       | You know what you're going to use, so you just do some simple
       | arithmetic and you've got a pretty accurate idea. Even with
       | serverless it's pretty easy. Tokens are so much harder,
       | especially when using it in a development setting. It's so hard
       | to have any reasonable forecast about how a team will use it, and
       | how many tokens will be consumed.
       | 
       | I'm starting to track my usage with a bit of a breakdown in the
       | hope that I'll find a somewhat reliable trend.
       | 
       | I suspect this is going to be one of the next big areas in cloud
       | FinOps.
        
         | prasoon2211 wrote:
         | This is partially why, at least for LLM-assisted coding
         | workloads, orgs are going with the $200 / mo Claude Code plans
         | and similar.
        
           | jsheard wrote:
           | Until the rug inevitably gets pulled on those as well. It's
           | not in your interest buy a $200/mo subscription unless you
           | use >$200 of tokens per month, and long term it's not in
           | their interest to sell you >$200 of tokens for a flat $200.
        
             | Hamuko wrote:
             | The pricing model works as long as people (on average)
             | think they need >$200 worth of tokens per month but
             | actually do something less, like $170/month. Is that
             | happening? No idea.
        
               | jsheard wrote:
               | Maybe that is what Anthropic is banking on, from what I
               | gather they obscure Max accounts actual token spend so
               | it's hard for subscribers to tell if they're getting
               | their moneys worth.
               | 
               | https://github.com/anthropics/claude-code/issues/1109
        
               | hombre_fatal wrote:
               | Well, the $200/mo plan model works as long as people on
               | the $100/mo plan is insufficient for some people which
               | works as long as the $17/mo plan is insufficient for some
               | people.
               | 
               | I don't see how it matters to you that you aren't
               | saturating your $200 plan. You have it because you hit
               | the limits of the $100/mo plan.
        
               | KallDrexx wrote:
               | I don't know about for people using CC on a regular
               | basis, but according to `ccusage`, I can trivially go
               | over $20 of API credits in a few days of hobby use. I'd
               | presume if you are paying for a $200 plan then you know
               | you have heavy usage and can easily exceed that.
        
               | jopsen wrote:
               | It's probably easier (and hence, cheaper) to finance the
               | AI infrastructure investments if you have a lot of
               | recurring subscriptions.
               | 
               | There is probably a lot of value in predictability.
               | Meaning it might be visible for a $200, to offer more
               | tokens than $200.
        
             | baq wrote:
             | meanwhile me hiding from accounting for spending $500 on
             | cursor max mode in a day
        
               | typpilol wrote:
               | Did you actually get 500 bucks worth of work out of it?
        
             | sellyme wrote:
             | > It's not in your interest buy a $200/mo subscription
             | unless you use >$200 of tokens per month
             | 
             | This is only true if you can find someone else selling them
             | at cost.
             | 
             | If a company has a product that cost them $150, but they
             | would ordinarily sell piecemeal for a total of $250,
             | getting a stable recurring purchase at $200 might be
             | worthwhile to them while still being a good deal for the
             | customer.
        
         | Spartan-S63 wrote:
         | > I suspect this is going to be one of the next big areas in
         | cloud FinOps.
         | 
         | It already is. There's been a lot of talk and development
         | around FinOps for AI and the challenges that come with that.
         | For companies, forecasting token usage and AI costs is non-
         | trivial for internal purposes. For external products, what's
         | the right unit economic? $/token, $/agentic execution, etc? The
         | former is detached from customer value, the latter is hard to
         | track and will have lots of variance.
         | 
         | With how variable output size can be (and input), it's a tricky
         | space to really get a grasp on at this point in time. It'll
         | become a solved problem, but right now, it's the Wild West.
        
         | mdasen wrote:
         | I agree that tokens are a really hard metric for people. I
         | think most people are used to getting something with a certain
         | amount of capacity per time and dealing with that. If you get a
         | server from AWS, you're getting a certain amount of capacity
         | per time. You still might not know what it's going to cost you
         | to do what you want - you might need more capacity to run your
         | website than you think. But you understand the units that are
         | being billed to you and it can't spiral out of control
         | (assuming you aren't using autoscaling or something).
         | 
         | When you get Claude Code's $20 plan, you get "around 45
         | messages every 5 hours". I don't really know what that means.
         | Does that mean I get 45 total conversations? Do minor followups
         | count against a message just as much as a long initial prompt?
         | Likewise, I don't know how many messages I'll use in a 5 hour
         | period. However, I do understand when I start bumping up
         | against limits. If I'm using it and start getting limited, I
         | understand that pretty quickly - in the same way that I might
         | understand a processor being slower and having to wait for
         | things.
         | 
         | With tokens, I might blow through a month's worth of tokens in
         | an afternoon. On one hand, it makes more sense to be flexible
         | for users. If I don't use tokens for the first 10 days, they
         | aren't lost. If I don't use Claude for the first 10 days, I
         | don't get 2,160 message credits banked up. Likewise, if I know
         | I'm going on vacation later, I can't use my Claude messages in
         | advance. But it's just a lot easier for humans to understand
         | bumping up against rate limits over a more finite period of
         | time and get an intuition for what they need to budget for.
        
           | Filligree wrote:
           | Both prefill and decode count against Claude's subscriptions;
           | your conversations are N^2 in conversation length.
           | 
           | My mental model is they're assigning some amount of API
           | credits to the account and billing the same way as if you
           | were using tokens, shutting off at an arbitrary point. The
           | point also appears to change based on load / time of day.
        
         | scuff3d wrote:
         | Also seems like a great idea to create a business models where
         | the companies aren't incentivised to provide the best product
         | possible. Instead they'll want to create a product just useful
         | enough to not drive away users, but just useless enough to temp
         | people to go up a tier, "I'm so close, just one more prompt and
         | it will be right this time!"
         | 
         | Edit: To be clear, I'm not talking about Zed. I'm talking about
         | the companies make the models.
        
           | potlee wrote:
           | While Apple is incentivized to ship a smaller battery to cut
           | costs, it is also incentivized to make their software
           | efficient as possible to make the best use of the battery
           | they do ship
        
             | scuff3d wrote:
             | That's not the same thing at all.
        
         | garrickvanburen wrote:
         | My rant on token-based pricing is primarily based on the
         | difficulty in consistently forecasting spend.....and also that
         | the ongoing value of a token is controlled by the
         | vendor...."the house always wins"
         | 
         | https://forstarters.substack.com/p/for-starters-59-on-credit...
        
           | coder543 wrote:
           | There are enough vendors that it's difficult for any one
           | vendor to charge too much per token. There are also a lot of
           | really good open-weight models that your business could self-
           | host if the hosted vendors _all_ conspire to charge too much
           | per token. (I believe it 's only economical to self-host big
           | models if you're using _a lot_ of tokens, so there is a
           | breakeven point.)
        
         | jklinger410 wrote:
         | Token based pricing works for the company, but not for the
         | user.
        
       | qsort wrote:
       | I wonder if first-party offerings like Codex and Claude will
       | follow suit. Most "agents" are utter nonsense, but they cooked
       | with the CLI tools. It'd be a shame to let go of them.
        
         | hashbig wrote:
         | Eventually that is the plan. Like we saw with Claude Code, they
         | want developers to get a taste of that unlimited and
         | unrestrained power of a state of the art model like Opus 4,
         | then slowly limit usage until you fully transition to metered
         | billing and deprecate subscription based billing.
        
       | prymitive wrote:
       | I can imagine the near future where companies "sponsor" open
       | source projects by donating tokens to "mine" a PR for a feature
       | they need.
        
         | ebrescia wrote:
         | I love this! Finally a more direct way for companies to sponsor
         | open source development. GitHub Sponsors helps, but it is often
         | so vague where the funding is going.
        
           | bsnnkv wrote:
           | More often than not, for individuals, it's barely
           | contributing to their living costs
        
           | scuff3d wrote:
           | If companies want to help they can just... I don't know...
           | give projects some money
        
           | drakythe wrote:
           | Unless companies also donate money to sponsor the code review
           | that will be required to be done by real human being I could
           | see this idea being a problem for maintainers. Yes you have
           | to code review a human being as well but a human being is
           | capable of learning and carrying that learning forward and
           | their next PR will be better, as well as being able to look
           | at past PRs to evaluate whether the user is a troll/bad actor
           | or someone who genuinely wants to assist with the project. An
           | LLM won't learn and will always spit out valid _looking_
           | code.
        
         | hombre_fatal wrote:
         | But the reason LLMs aren't used to build features isn't because
         | they are expensive.
         | 
         | The hard work is the high level stuff like deciding on the
         | scope of the project, how it should fit in to the project, what
         | kind of extensibility the feature might need to be built with,
         | what kind of other components can be extended to support it,
         | (and more), and then reviewing all the work that was done.
        
       | sharkjacobs wrote:
       | This whole business model of trying to shave off or arbitrage a
       | fraction of the money going to OpenAI and Anthropic just sucks.
       | And it seems precarious. There's no honest way to resell tokens
       | at a profit, and everyone knows it.
        
         | Havoc wrote:
         | >There's no honest way to resell tokens at a profit, and
         | everyone knows it.
         | 
         | Agree with the sentiment, but I do think there are edge cases.
         | 
         | e.g. I could see a place like openrouter getting away with a
         | tiny fractional markup based on the value they provide in the
         | form of having all providers in one place
        
           | Lalabadie wrote:
           | The issue with a model like this (fixed small percentage) is
           | that your biggest clients are the most incentivized to move
           | away.
           | 
           | At scale, OpenRouter will instead get you the lower high-
           | volume fees they themselves get from their different
           | providers.
        
         | thelastbender12 wrote:
         | Sorry, how is this new pricing anything but honest? They
         | provide an editor you can use to - optimize the context you
         | send to the LLM services - interact with the output that comes
         | out of them
         | 
         | Why does not justify charging a fraction of your spend on the
         | LLM platform? This is pretty much how every service business
         | operates.
        
           | hu3 wrote:
           | There's now greater incentive for Zed to stuff more content
           | in the prompts to inflate tokens used and thus profit more.
           | Or at least be less zealous.
           | 
           | This is not a new concern. And is not unique to Zed.
        
         | drakythe wrote:
         | For companies where that is their entire business model I
         | absolutely agree. Zed is a solid editor with additional LLM
         | integration features though, so this move would seem to me to
         | just cover their costs + some LLM integration development
         | funds. If their users don't want to use the LLM then no skin
         | off Zed's back unless they've signed some guaranteed usage
         | contract.
        
         | dinvlad wrote:
         | The whole business model even for OAI/Anthropic is
         | unsustainable.. they are already running it at a huge loss atm,
         | and will do for the foreseeable future. The economics simply
         | doesn't work, unfortunately or not
        
       | andrewmcwatters wrote:
       | Am I wrong in that GitHub Copilot Pro apparently has the best
       | overall token spend when considering agentic editors?
        
         | ramon156 wrote:
         | Better than Gemini Pro 2.5? Github Copilot doesn't even support
         | tooling in Zed yet. It's been months..
        
         | hu3 wrote:
         | $10 GitHub Copilot Pro plan works for me in VSCode.
         | 
         | I've been exclusively using Claude Sonnet 4 model in VSCode and
         | so far I've used 90% of the premium quota at the end of the
         | months. I can always use GPT4.1 or GPT5-mini for free if need
         | be.
        
           | andrewmcwatters wrote:
           | Yeah, that's what I've been using since its release as well.
           | I don't really see a point in trying the competition. It
           | can't be better than this.
        
       | genshii wrote:
       | I'm personally looking forward to this change because I currently
       | pay $20/month just to get edit prediction. I use Claude Code in
       | my terminal for everything else. I do wish I could _just_ pay for
       | edit prediction at an even lower price, but I can understand why
       | that 's not an option.
       | 
       | I'm curious if they have plans to improve edit prediction though.
       | It's honestly kind of garbage compared to Cursor, and I don't
       | think I'm being hyperbolic by calling it garbage. Most of the
       | time it's suggestions aren't helpful, but the 10-20% of the time
       | it _is_ helpful is worth the cost of the subscription for me.
        
         | morgankrey wrote:
         | We have a significant investment underway in edit predictions.
         | We hear you, more soon.
        
           | genshii wrote:
           | That's great to hear, thanks!
        
           | sippeangelo wrote:
           | This is the one thing keeping me from switching from Cursor.
           | I much prefer Zed in every other way. Exciting!
        
             | hombre_fatal wrote:
             | Yeah, Cursor tab completion is basically in the realm of
             | magical mind reading and might still be the most insane
             | productivity demonstration of LLM tech in software.
             | 
             | It obsoleted making Vim macros and multiline editing for
             | example. Now you just make one change and the LLM can
             | derive the rest; you just press tab.
             | 
             | It's interesting that the Cursor team's first iteration is
             | still better than anything I've seen in their competitors.
             | It's been an amazing moat for a year(?) now.
        
               | sippeangelo wrote:
               | I agree. I wish they focused more on it. I'd love to be
               | able to give it a few sentences of instructions to make
               | it even more effective for me. It's so much more of a
               | productivity boon than all the coding agent stuff ever
               | was.
        
               | pdntspa wrote:
               | I could say the same about the AI-assisted autocomplete
               | in IDEA. Wonder how they compare...
        
           | pkilgore wrote:
           | This is very very exciting.
        
         | inerte wrote:
         | That's been my workflow also. Claude Code / OpenAI Codex most
         | of the time, when I have to edit files Cursor's auto-complete
         | is totally worth the $20.
        
         | chewz wrote:
         | I have never used Zed predictions but $20 for 500 prompts is
         | quite a good deal. I use it mostly with Opus for some hard
         | cases.
        
           | typpilol wrote:
           | 10 bucks on copilot and you get unlimited + unlimited gpt4.1
           | etc
           | 
           | Copilot is the best value by far
        
       | okokwhatever wrote:
       | This is going to be a blood bath for many freelancers if the
       | trend continues with other platforms. Mark my words.
        
       | WD-42 wrote:
       | Good change. I'm not a vibe coder, I use Zed Pro llm integration
       | more like glorified stack overflow. I value Zed more for being an
       | amazing editor for the code I actually write and understand.
       | 
       | I suspect I'm not alone on this. Zed is not the editor for
       | hardcore agentic editing and that's fine. I will probably save
       | money on this transition while continuing to support this great
       | editor for what it truly shines at: editing source code.
        
       | AbuAssar wrote:
       | Zed and Warp were two promising Rust-based projects that I
       | closely monitor. Currently, both projects are progressing towards
       | becoming a generic AI Agentic code platform.
        
         | scuff3d wrote:
         | Until now I've never really come across a comment on Hackernews
         | I thought was AI generated...
        
           | AbuAssar wrote:
           | You are partially right, after I wrote the comment I used the
           | writing tools in macOS to rewrite it in a professional tone.
           | 
           | The wording may sound AI generated but the gist of the
           | comment is my true opinion
        
       | muratsu wrote:
       | For those of us building agentic tools that require similar
       | pricing, how does one implement it? OpenRouter seems good for the
       | MVP, but I'm curious if there are alternatives down the line.
        
       | VGHN7XDuOXPAzol wrote:
       | > Token-agnostic prompt structures obscure the cost and are rife
       | with misaligned incentives
       | 
       | Saying that, token-based pricing has misaligned incentives as
       | well: as the editor developer (charging a margin over the number
       | of tokens) or AI provider, you benefit from more verbose input
       | fed to the LLMs and of course more verbose output from the LLMs.
       | 
       | Not that I'm really surprised by the announcement though, it was
       | somewhat obviously unsustainable
        
       | dmix wrote:
       | I just asked this exact question about Zed pricing 2 days ago
       | 
       | https://news.ycombinator.com/item?id=45333425
        
       | dinobones wrote:
       | Making this prediction now, LLM pricing will eventually be priced
       | in bytes.
       | 
       | Why: LLMs are increasingly becoming multimodal, so an image
       | "token" or video "token" is not as simple as a text token. Also,
       | it's difficult to compare across competitors because tokenization
       | is different.
       | 
       | Eventually prices will just be in $/Mb of data processed. Just
       | like bandwidth. I'm surprised this hasn't already happened.
        
         | vtail wrote:
         | Hm... why not tokens as reported by each LLM provider? They
         | already handle pricing for images etc.
        
         | jermaustin1 wrote:
         | The problem is that tokens don't all equate to the same size. A
         | megabyte of some random json is a LOT more tokens than a
         | megabyte of "Moby Dick".
        
         | dragonwriter wrote:
         | > Why: LLMs are increasingly becoming multimodal, so an image
         | "token" or video "token" is not as simple as a text token.
         | 
         | For autoregressive token-based multimodal models, image tokens
         | are as straightforward as text tokens, and there is no reason
         | video tokens wouldn't also be. (If models also switch
         | architecture and multimodal diffusion models, say, become more
         | common, then, sure, a different pricing model more tied to
         | actual compute cost drivers for that architecture are likely
         | but... even that isn't likely to be bytes.)
         | 
         | > Also, it's difficult to compare across competitors because
         | tokenization is different.
         | 
         | That's a reason for incumbents to prefer _not_ to switch,
         | though, not a reason for them to switch.
         | 
         | > Eventually prices will just be in $/Mb of data processed.
         | 
         | More likely they would be in floatint point operations expended
         | processing them, but using tokens (which are the primary
         | drivers for the current LLM architectures) will probably
         | continue as long as the architecture itself is doninant.
        
           | oblio wrote:
           | > For autoregressive token-based multimodal models, image
           | tokens are as straightforward as text tokens, and there is no
           | reason video tokens wouldn't also be.
           | 
           | In classical computing, there is a clear hierarchy: text <
           | images <<< video.
           | 
           | Is there a reason why video computing using LLMs shouldn't be
           | much more intensive and therefore costly than text or image
           | output?
        
             | Filligree wrote:
             | Of course it's more expensive. It's still tokens, but
             | considerably more of them.
        
               | oblio wrote:
               | That's the thing, I can't visualize (and I don't think
               | most people can) what "tokens" represent for image or
               | video outputs.
               | 
               | For text I just assume them to be word stems or more like
               | work-family-members (cat-feline-etc).
               | 
               | For images and videos I guess each character, creature,
               | idea in it is a token? Blue sky, cat walking around,
               | gentleman with a top hat, multiplied by the number of
               | frames?
        
               | dragonwriter wrote:
               | > For images and videos I guess each character, creature,
               | idea in it is a token?
               | 
               | No, for images, tokens would, I expect, usually be
               | asymptotically proportional to the area of the image
               | (this is certainly the case with input token for OpenAIs
               | models that take image inputs; outputs are more opaque);
               | you probably won't have a neat one-to-one intuition for
               | what one token represents, but you don't need that for it
               | to be useful and straightforward for understanding
               | pricing, since the mathematical relationship of tokens to
               | size can be published and the size of the image is a
               | known quantity. (And videos conceptually could be like
               | images with an additional dimension.)
        
             | dragonwriter wrote:
             | No, it'll certainly be more expensive in any conceivable
             | model that handles all three modalities, but if the model
             | uses an architecture like current autoregressive, token-
             | based multimodal LLMs/VLMs, tokens will make just as much
             | sense as the basis for pricing, and be similarly
             | straightforward, as with text and images.
        
           | efskap wrote:
           | To clarify, "as straightforward" = same dimensionality? I
           | guess it would have to be, to be usable in the same embedding
           | space.
        
         | jstummbillig wrote:
         | Why this instead of cpu/gpu time?
        
           | typpilol wrote:
           | I'm assuming token count and usage is pretty closely tied
        
           | antiframe wrote:
           | CPU/GPU time is opaque to me before I send my data, but
           | tokens I can count before I decide to send it. That means I
           | can verify the metering. With CPU time I send the data and
           | then the company says "That cost X CPU units, which is $500".
        
         | mhuffman wrote:
         | No one is going to give up token-based pricing. The main
         | players can twiddle their models to make anything any amount of
         | tokens they choose.
        
       | vtail wrote:
       | Prediction: the only remaining providers of AI-assisted tools in
       | a few years will be the LLM companies themselves (think claude
       | code, codex, gemini, future xai/Alibaba/etc.), via CLIs +
       | integrations such as ASP.
       | 
       | There is very little value that a company that _has_ to support
       | multiple different providers, such as Cursor, can offer on top of
       | tailored agents (and  "unlimited" subscription models) by LLM
       | providers.
        
         | computerex wrote:
         | I don't know. Foundation models are very good, and you can get
         | a surprising amount of mileage from them by using them with low
         | level interfaces. But personally I think companies building
         | development tools of the future will use LLMs to build systems
         | with increasing capabilities. I think a lot of engineering
         | challenges remain in scaling LLM's to take over day to day in
         | programming, and the current tools are scratching the surface
         | of what's possible when you combine LLMs with traditional
         | systems engineering.
        
         | serbuvlad wrote:
         | I recently started using Codex (OpenAI's Claude Code) and it
         | has a VSCode extension that works like a charm. I tried out
         | Windsurf a while ago. And the Codex extension simply does
         | everything that Windsurf did. I guess it doesn't show changes
         | at well, (it shows diffs in it's own window instead of in the
         | file), but I can just check a git diff graphically (current
         | state vs. HEAD) if I really wanted that.
         | 
         | I am really tempted to buy ChatGPT Pro, and probably would have
         | if I lived in a richer country (unfortunetley purchase power
         | parity doesn't equalize for tech products). The problem with
         | Windsurf (and presumably Cursor and others) is that you buy the
         | IDE subscription and then still have to worry about usage
         | costs. With Codex/Claude Code etc., yeah, it's expensive, but,
         | as long as you're within the usage limits, which are hopefully
         | reasonable for the most expensive prices, you don't have to
         | worry about it. AND you get the web and phone apps with GPT 5
         | Pro, etc.
        
         | rudedogg wrote:
         | If you look at even the Claude/OpenAI chat UIs, they kind of
         | suck. Not sure why you think someone else can't/won't do it
         | better. Yes, the big players will copy what they can, but they
         | also need to chase insane growth and getting every human on
         | earth paying for an LLM subscription.
         | 
         | A tool that is good for everyone is great for no one.
         | 
         | Also, I think we're seeing the limits on "value" of a chat
         | interface already. Now they're all chasing developers since
         | there's a real potential to improve productivity (or sadly cut-
         | costs) there. But even that is proving difficult.
        
       | oakesm9 wrote:
       | I completely get why this pricing is needed and it seems fair.
       | There's a major flaw in the announcement though.
       | 
       | I get that the pro plan has $5 of tokens and the pricing page
       | says that a token is roughly 3-4 characters. However, it is not
       | clear:
       | 
       | - Are tokens input characters, output characters, or both?
       | 
       | - What does a token cost? I get that the pricing page says it
       | varies by model and is " API list price +10%", but nowhere does
       | it say what these API list prices are. Am I meant to go to The
       | OpenAI, Anthropic, and other websites to get that pricing
       | information? Shouldn't that be in a table on that page which each
       | hosted model listed?
       | 
       | --
       | 
       | I'm only a very casual user of AI tools so maybe this is clear to
       | people deep in this world, but it's not clear to me just based on
       | Zelda pricing page exactly how far $5 per month will get me.
        
         | morgankrey wrote:
         | List here: https://zed.dev/docs/ai/models. Thanks for the
         | feedback, we'll make sure this is linked from the pricing page.
         | Think it got lost in the launch shuffle.
        
           | oakesm9 wrote:
           | All makes sense. I presumed it was an oversight.
           | 
           | It's hard for me to conceptualise what a million tokens
           | actually looks like, but I don't think there's a way around
           | that aside from making proving some concrete examples of
           | inputs, outputs, and the number of tokens that actually is. I
           | guess it would become clearer after using it a bit.
        
             | morgankrey wrote:
             | Now live: https://zed.dev/pricing#what-is-a-token. Thanks
             | for the feedback
        
       | bananapub wrote:
       | seems fine - they're aligning their prices with their costs.
       | 
       | presumably everyone is just aiming or hoping for inference costs
       | to go down so much that they can do a unlimited-with-tos like
       | most home Internet access etc, because this intermediate phase of
       | having to count your pennies to ask the matrix multiplier
       | questions isn't going to be very enjoyable or stable or encourage
       | good companies to succeed.
        
       | giancarlostoro wrote:
       | I was just thinking this morning about how I think Zed should
       | rethink their subscription because its a bit pricey if they're
       | going to let you just use Claude Code. I am in the process of
       | trying out Claude and figured just going to them for the
       | subscriptions makes more sense.
       | 
       | I think Zed had a lot of good concepts where they could make paid
       | AI benefits optional longer term. I like that you can join your
       | devs to look at different code files and discuss them. I might
       | still pay for Zed's subscription in order to support them long
       | term regardless.
       | 
       | I'm still upset so many hosted models dont just let you use your
       | subscription on things like Zed or JetBrains AI, what's the point
       | of a monthly subscription if I can only use your LLM in a
       | browser?
        
         | hamandcheese wrote:
         | > I'm still upset so many hosted models dont just let you use
         | your subscription on things like Zed or JetBrains AI, what's
         | the point of a monthly subscription if I can only use your LLM
         | in a browser?
         | 
         | This is her another reason why CLI-based coding agents will
         | win. Every editor out there trying to be the middle man between
         | you and an AI provider is nuts.
        
           | oblio wrote:
           | Wouldn't the last step just be an API? That would allow
           | direct integration from everywhere.
        
             | praseodym wrote:
             | There is one, developed by the Zed team in collaboration
             | with Gemini. And Claude Code is also supported now.
             | 
             | https://agentclientprotocol.com/overview/introduction
        
         | ibejoeb wrote:
         | > if they're going to let you just use Claude Code
         | 
         | I'm pretty sure that's only while it's in preview, just like
         | they were giving away model access before that was formally
         | launched. Get it while it's hot.
        
       | dinvlad wrote:
       | Another one bites the dust :-( I hope at least Windsurf stays the
       | same..
        
       | pkilgore wrote:
       | This is much better for me but I really want a plan that includes
       | zero AI other than edit prediction and BYOK for the rest.
       | 
       | But as a mostly claude max + zed user happy to see my costs go
       | down.
        
       | blutoot wrote:
       | Why is most of the AI-tooling industry still stuck on this "bring
       | your own key" model?
        
         | sethhochberg wrote:
         | What would you propose as an alternative?
         | 
         | As a corporate purchaser, "bring your own key" is just about
         | the only way we can allow our employees to stay close to the
         | latest happenings in a rapidly moving corner of the industry.
         | 
         | We need to have a decent amount of trust in the model execution
         | environment and we don't like having tons of variable-cost
         | subscriptions. We have that trust in our corporate-managed
         | OpenAI tenant and have good governance and budget controls
         | there, so BYOK lets us have flexibility to put different
         | frontends in front of our trusted execution environment for
         | different use cases.
        
         | Macha wrote:
         | The companies actually providing the models charge by token and
         | this lets the tooling avoid having to do cost planning for
         | something with a bunch of unknowns and push the risk of
         | overspend to customers.
        
       | hmokiguess wrote:
       | Zed was supposed to be the answer to Atom / Sublime Text in my
       | opinion, and I kinda do want to use it as my main driver, but it
       | just isn't there yet for me. It's shameful because I like its
       | aesthetics as a product more than the competition out there.
       | 
       | Just this other day I tried using it for something it sort of
       | advertised itself as the superior thing, which was to load this
       | giant text file I had instantly and let me work on it.
       | 
       | I then tried opening this 1GB text file to do a simple
       | find/replace on it only to find macOS run out of system memory
       | with Zed quickly using 20gb of memory for that search operation.
       | 
       | I then switched to vscode, which, granted opened it in a buffered
       | sort of way and limited capability, but got the job done.
       | 
       | Maybe that was a me issue I don't know, but aside from this one-
       | off, it doesn't have a good extensions support in the community
       | for my needs yet. I hope it gets there!
        
         | j_bum wrote:
         | VSCode is my go to for large text file interaction on macOS.
         | 
         | TextEdit may be worth looking into as well? Haven't tested it
         | for large files before.
        
           | dewey wrote:
           | I have Sublime Text installed for the onlu use case of
           | opening large files. Nothing comes close.
        
             | CharlesW wrote:
             | Googling around a bit, Sublime Text doesn't seem to be
             | particularly good at this:
             | https://forum.sublimetext.com/t/unable-to-open-a-large-
             | text-...
             | 
             | In my experience, BBEdit will open files that kill other
             | editors: "Handling large files presents no intrinsic
             | problems for BBEdit, though some specific operations may be
             | limited when dealing with files over 2GB in size."
        
               | xrisk wrote:
               | While I don't know if the claim is true, you've linked a
               | post from 2012...
        
               | roto wrote:
               | I have always found sublime to be the best at large
               | files, well over 1gb. Since you mention bbedit, maybe
               | this is some mac specific issue? I really don't know. But
               | at least among people i know, opening large files has
               | effectively become its main USP.
               | 
               | Should be noted that the linked post is almost 15 years
               | old at this point too, so perhaps not the most up to date
               | either.
        
           | hmokiguess wrote:
           | Speaking of TextEdit, I like what the folks at CodeEdit are
           | doing. They are moving slow and focusing on just the core
           | parts. Maybe I should go give them a try too!
        
           | typpilol wrote:
           | Vscode has a special optimizations in place for large files.
           | That's why it works so good.
           | 
           | You can actually disable it in the settings if you want it to
           | try and render the entire thing at once
        
         | laweijfmvo wrote:
         | Similar experience: I added a folder to my zed project that was
         | too big, causing zed to lock up and eventually crash. But
         | because the default setting was to re-open projects on launch,
         | I was stuck in a loop where I couldn't remove the folder
         | either. Eventually found a way to clear the recent projects,
         | load an empty editor, and change the setting to avoid it in the
         | future.
        
           | mr90210 wrote:
           | Ok, how big was your project?
           | 
           | My JetBrains IDEs (RustRover, Goland) probably would have
           | choked out too.
        
             | tux3 wrote:
             | You can open large codebases in Jetbrains IDEs and it takes
             | forever to index, but it shouldn't outright crash or
             | completely freeze.
             | 
             | You can open the kernel in CLion. Don't expect the advanced
             | refactoring features to work, but it can deal with a ~40
             | million lines project folder for example
        
           | silverwind wrote:
           | Big files/projects is where Sublime really shines. I hope Zed
           | can replicate that performance.
        
         | klaussilveira wrote:
         | I feel like Zed stopped working on the editor itself since AI
         | was rolled out. I also wanted it to be the open-source
         | alternative to Sublime, but nothing comes close.
        
           | hmokiguess wrote:
           | Yeah exactly this! I get they want to stay in the game and
           | follow the market, but I'm sad they're not being more
           | aggressive on that original vision. I still think there could
           | be a huge payoff for them if they invested more on their
           | brand and aesthetics of a more polished and comfy editor.
           | 
           | The way I see it, we're sort of living in a world where UX is
           | king. (Looking at you Cursor)
           | 
           | I feel like there's a general sentiment where folks just want
           | a sense of home with their tools more than anything. Yes they
           | need to work, but they also need to work for you in your way.
           | Cursor reinvented autocomplete with AI and that felt like
           | home for most, what's next? I see so much focus on Agents but
           | to me personally that feels more like it should live on the
           | CI/CD layer of things. Editors are built for humans,
           | something isn't quite there yet, excited to see how it
           | unfolds.
        
           | bombcar wrote:
           | If the full magnitude of products that stopped working on the
           | main product and started to try to shoehorn AI in was known,
           | the economy would collapse overnight.
        
             | kace91 wrote:
             | It really is concerning. I keep an excel sheet with links
             | of all companies I could apply to whenever i change jobs,
             | and checking it the other day practically every row was now
             | selling an ai product.
        
             | 0x1ceb00da wrote:
             | Intellij wasn't immune to it either. Number of bugs has
             | exploded since they started adding ai features to their
             | ide, none of which I used for more than 5 minutes.
        
           | hakanensari wrote:
           | I think their recent push to delegate to CLI agents in the
           | agent panel is the right direction. Claude Code has been
           | running in Zed for the past month. Sure, there are SDK
           | limitations and kinks to iron out, but it's moving quickly.
           | I'm into it.
        
             | hmokiguess wrote:
             | I get what you are saying, and I think they are doing a
             | good job there as well. That said, it still feels like
             | something is missing in that whole workflow to me.
             | 
             | I sometimes worry if we are moving too fast for no reason.
             | Some things are becoming standards in an organic way but
             | they feel suboptimal in my own little bias bubble corner.
             | 
             | Maybe I am getting old and struggling to adapt to the new
             | generation way of getting work done, but I have a gut
             | feeling that we need to revisit some of this stuff more
             | deliberately.
             | 
             | I still see Agents as something that will be more like a
             | background thread that yields rather than a first class
             | citizen inside the Editor you observe as it goes.
             | 
             | I don't know about you, but I feel an existential dread
             | whenever I prompt an Agent and turn into a vegetable
             | watching it breathe. -- am I using it wrong? Should I be
             | leaving and coming back later? Should I pick a different
             | file and task while it's doing its thing?
        
           | jamesgeck0 wrote:
           | There have been improvements recently, but it still has some
           | of the worst text rendering of any editor on macOS, if you
           | have a non-4K display plugged in. Rendering text is kind of a
           | big deal!
        
         | mr90210 wrote:
         | Speaking of aesthetics, I switched back to VSCode but I ended
         | up installing the theme "Zed One Theme" and switched the
         | editor's font to "IBM Plex Mono".
         | 
         | I know it's not Zed, but I am pretty satisfied with the
         | results.
        
         | shafyy wrote:
         | Zed is the opposite of Sublime. Zed is VC funded and will
         | eventually be enshittified. Sublime is not and has been going
         | strong for many years.
        
           | hmokiguess wrote:
           | Yeah fair point. I think CodeEdit is a perhaps a closer
           | comparison there
        
         | jeltz wrote:
         | My experience with Zed was that it was a buggy mess with
         | strange defaults. I had high hopes until I tried it but it was
         | not even close to working for my pretty normal use case of
         | coding C so I went back to Sublime.
        
           | jimmydoe wrote:
           | same.
           | 
           | I reported some issue a few months back but saw they have
           | thousands to deal with so mine will understandably stay open
           | forever.
        
         | indigodaddy wrote:
         | CotEditor is quite good with very large files.
         | 
         | https://github.com/coteditor/CotEditor
        
         | epolanski wrote:
         | They don't care about the editor anymore, it's all about
         | becoming millionaires.
        
       | bitwize wrote:
       | Thought this was 2015 for a sec and this was about Zed Shaw.
        
       | acaloiar wrote:
       | Tokens are an implementation detail that have no business being
       | part of product pricing.
       | 
       | It's deliberate obfuscation. First, there's the simple math of
       | converting tokens to dollars. This is easy enough; people are
       | familiar with "credits". Credits can be obfuscation, but at least
       | they're honest. The second and more difficult obfuscation to
       | untangle is how one converts "tokens" to "value".
       | 
       | When the value customers receive from tokens slips, they pay the
       | same price for the service. But generative AI companies are under
       | no obligation to refund anything, because the customer paid for
       | tokens, and they got tokens in return. Customers have to trust
       | that they're being given the highest quality tokens the provider
       | can generate. I don't have that trust.
       | 
       | Additionally, they have to trust that generative AI companies
       | aren't padding results with superfluous tokens to hit revenue
       | targets. We've all seen how much fluff is in default LLM
       | responses.
       | 
       | Pinky promises don't make for healthy business relationships.
        
         | mrcwinn wrote:
         | Doesn't prompt pricing obfuscate token costs by definition? I
         | guess the alternative is everyone pays $500/mo. (And you'd
         | still get more value than that.)
        
         | ibejoeb wrote:
         | Tokens aren't that much more opaque than RAM GB/s for functions
         | or whatever. You'd have to know the entire infra stack to
         | really understand it. I don't really have a suggestion for a
         | better unit for that kind of stuff.
        
       | travisgriggs wrote:
       | I love this for Zed. I hate that I'm going to have deal with the
       | model providers more directly. Because I don't know what tokens
       | are.
       | 
       | It's like if McDonalds changed their pricing model to be some
       | complex formula involving nutrition properties (calories, carbs,
       | etc) plus some other things (carbon tax, local state taxes,
       | whatever) and moved to a pay as you bite model. You start eating
       | a Big Mac, but every bite has different content proportions, so
       | every bite is changing in price as you go. Only through trial and
       | error would you figure out how to eat. And the fact that the
       | "complex formula" is prone to real time change at any point,
       | makes it impossible to get excited about eating.
        
       | travisgriggs wrote:
       | I really want to be able to see outline panel, project panel,
       | agent panel, and terminal panel all at the same time. If I stay
       | at $20/mo, can you guys fix the please?
       | 
       | I'm tired of toggling between them.
        
       | rdtsc wrote:
       | I am wondering why they couldn't have foreseen this. Was it
       | really a failure to predict the need to charge for tokens
       | eventually, or was planned from the start that way -- get people
       | to use the unlimited option for a bit, they get hooked, then
       | switch them to per-token subscriptions.
        
       | indymike wrote:
       | I'm going through this with pricing for our product and the
       | battle is:
       | 
       | Customer: wants predictable spend Sales/Marketing: wants an
       | amazing product that is easy to sell (read does the mostest for
       | the leastest) Advisors: want to see the SaaS model in all it's
       | glory (i.e. 85% margin primarily from oversubscription of
       | infrastructure) Finance: wants to make money and is really
       | worried about variable costs ruining SaaS profit margin
       | 
       | A couple of thoughts:
       | 
       | 1. AI-based cost is mostly variable. Also, things like messaging,
       | phone minutes and so on are variable. Cloud expenses are also
       | variable... There's a theme and it's different.
       | 
       | 2. The unit of value most ai software is delivering is work that
       | is being done by people or should be being done and is not.
       | 
       | 3. Seems like it's time to make friends with a way to make money
       | other than subscription pricing.
        
       | nextworddev wrote:
       | The first thing they do after getting funded by Sequoia lmao
        
       | marcopolo wrote:
       | I'm glad to see this change. I didn't much use the AI features,
       | but I did want to support Zed. $20 seemed a bit high for that
       | signal. $10 seems right. $5 with no tokens would be nicer.
       | 
       | Great work folks
        
       | hendersoon wrote:
       | So they're essentially charging $5/month for unlimited tab
       | completions, when you get 2k for free. That seems reasonable,
       | many could just not pay anything at all.
       | 
       | But in the paid plan they charge 10% over API prices for metered
       | usage... and also support bring your own API. Why would anyone
       | pay their +10%, just to be nice?
       | 
       | This is the same problem cursor and windsurf are facing. How the
       | heck do you make money when you're competing with cline/roocode
       | for users who are by definition technically sophisticated? What
       | can you offer that's so wonderful that they can't?
        
       | hit8run wrote:
       | Say what you want but Sublime is the GOAT.
        
       | ibejoeb wrote:
       | > Additional usage billed at API list price +10%
       | 
       | I'm hoping that zed is someday able to get discount bulk pricing
       | to resell through their providers.
       | 
       | With the 10% markup, is there any benefit to using zed's provider
       | vs. BYOK?
        
       ___________________________________________________________________
       (page generated 2025-09-24 23:01 UTC)