Side note: I don’t know what Anthropic changed but now Claude Code consumes the quota incredibly fast. I have the Max5 plan, and it just consumed about 10% of the session quota in 10 minutes on a single prompt. For $100/month, I have higher expectations.
The opacity isn't accidental. When customers can't forecast cost, they can't build an internal business case to switch, which is more durable than lock-in from product quality. The first AI coding vendor to offer deterministic $/task pricing will own enterprise procurement conversations. Right now the entire market is hiding behind 'tokens' specifically to avoid that comparison.
I completely agree that requests are what should be charged for. But I think there are two things, given that requests aren't all going to cost the same amount:
1. Estimate free invoicing the requests and letting users figure it out after the fact.
2. Somehow estimating cost and telling users how much a request will cost.
I had to double check that they'd removed the non-1M option, and... WTF? This is what's in `/config` → `model`
1. Default (recommended) Opus 4.6 with 1M context · Most capable for complex work
2. Sonnet Sonnet 4.6 · Best for everyday tasks
3. Sonnet (1M context) Sonnet 4.6 with 1M context · Billed as extra usage · $3/$15 per Mtok
4. Haiku Haiku 4.5 · Fastest for quick answers
So there's an option to use non-1M Sonnet, but not non-1M Opus?
Except wait, I guess that actually makes sense, because it says Sonnet 1M is billed as extra usage... but also WTF, why is Sonnet 1M billed as extra usage? So Opus 1M is included in Max, but if you want the worse model with that much context, you have to pay extra? Why the heck would anyone do that?
The screen does also say "For other/previous model names, specify with --model", so I assume you can use that to get 200K Opus, but I'm very confused why Anthropic wouldn't include that in the list of options.
What a strange UX decision. I'm not personally annoyed, I just think it's bizarre.
Thanks. I quickly burned through $100 in credit when I started using Opus 4.6 in OpenCode via OpenRouter. My session stopped and was getting an error not representative of credit availability, so was surprised after a few minutes when I finally realized Opus just destroyed those credits on a bullshit reasoning loop it got stuck in. Anthropic seems to know that the expanded context is better for their bottom line as they've defaulted it now.
And as others have said it's very easy to burn token usage on the $100/month plan. It's getting to the point where it's going to very much make sense to do model routing when using coding tooling.
Anthropic is not building good will as a consumer brand. They've got the best product right now but there's a spring charging behind me ready to launch me into OpenCode as soon as the time is right.
I'd like to use Opus with OpenCode right now to combine the best TUI agent app with the best LLM. But my understanding is Anthropic will nuke me from orbit if I try that.
You can use Opus with OpenCode anytime you want, just not with the Claude plan. You can use it via API with any provider, including Anthropic's API. You can use it with Github Copilot's plan. The only thing you can't do without getting banned is use OpenCode with one of Claude's plans.
I've heard this a few times lately, but this past weekend I built a website for a friend's birthday, and it took me several hours and many queries to get through my regular paid plan. I just use default settings (Sonnet 4.6, medium effort, thinking on).
I'm guessing Opus eats up usage much, much faster. I don't know what's going on, since a lot of people are hitting limits and I don't seem to be.
I waited until off peak hours to use Opus 4.6 to do some research. One prompt consumed 100% of my 5h limit and 15% of my weekly usage. Even off peak it's still insane. Opus didn't even manage to finish what it was doing.
Even with Opus I don’t usually hit limits on the standard plan. But I am not doing professional work at the moment and I actually alternate between using the LLM and reading/writing code the old fashioned way. I can see how you’d blow through the quota quickly if you try to use LLMs as universal problem solvers.
I've been jumping from Claude -> Gemini -> GPT Codex. Both Claude and Gemini really reduced quotas and so I cancelled. Only subbed GPT for the special 2x quota in March and now my allocation is done as well.
I decided to give opencode go a try today. It's $5 for the first month. Didn't get much success with Kimi K2, overly chatty, built too complex solutions - burned 40% of my allocation and nothing worked. ¯\_(ツ)_/¯.
But Minimax m2.7. Wow, it feels just like Claude Opus 4.6. Really has serious chops in Rust.
Tomorrow/Wednesday will try a month of their $40 plan and see how it goes.
Reminds me of when I would mess with my friends on "pay per text" plans by sending them 10 text messages instead of just 1. I should start paying attention to unattended laptops and blow up some token usage in the same manner.
I responded with a mix of mostly B and C answers and got “advanced.” Yet, as pointed out by another commenter, selecting all D answers (which would make you an expert!) gets you called a beginner.
I can only assume the quiz itself was vibe-coded and not tested. What an incredible time we live in.
Or that it's taking into account the Dunning-Kruger effect. In that, if you think you are an expert in all cases, you are really a beginner in everything.
I'm a beginner with agentic coding. I vibe code something most days, from a few lines up to refactors over a few files. I don't knowingly use skills, rarely _choose_ to call out to tools, haven't written any skills and only one or two ad hoc scripts, and have barely touched MCPs (because the few I've used seem flaky and erratic). I answered as such and got... intermediate.
Strongly agree with the sentiment, but I'd say if you're familiar with the terminal you may as well just install it and truly 'learn by doing'!
I could see this being great for true beginners, but for them it might be nice to have even some more basics to start (how do I open the terminal, what is a command, etc).
I love the pedagogical approach here and the ability to easily hone in on your level before diving into content. Your approach would work really well for other subjects as well.
Why wpuld anyone want to "learn" how to use some non-deterministic black box of bullshit that is frequently wrong? When you get different output fkr the same input, how do you learn?
How is that beneficial? Why would you waste your time learning something that is frequently changing at the whims of some greedy third party? No thanks.
there is certainly a future where this isn't the case. Learning how to use AI and use it in your workflows will likely for sure be a part of any serious dev's future, but being beholden to a data center does not seem to reflect reality. Consider all the 5m-8m models and how powerful they are today compared to what the best models did 2 years ago. If you want to stay absolute bleeding edge model wise, sure you'll be stuck at a data center for some time...
Why isn't this just kinda seen as a repeat of the original birth of computers? Consider the IBM 350 (3.5mb) rented in the 50s for thousands per month. Now I have a drawer filled with SD cards that go up to 128gb that i cant even give away.
I think it's funny and interesting how LLMs are commoditizing information generation. It's completely expected, but also somewhat challenging to figure out what the best combination of "learning" "fact" systems is.
I'd be curious to know more about how this compares to other approaches.
Are people again learning a new set of tools? Just tell the AI what you want, if the AI tool doesn't allow that then tell another Ai tool to make you a translation layer that will convert the natural language to the commands etc. What's the point of learning yet another tool?
I haven't used Claude, but the problem seems to be not refusal, but cheerful failure. "Sure, I'll help you with that!" And it produces something wrong in obvious and/or subtle ways.
That's like saying there's no point in attending a lecture on "how to get the best out of your time at University" because University courses are taught in spoken language so you could just ask the professors.
The idea that AI can write code like a seasoned software developer but not being able to use its own tooling that can be learned through 11 chapters tutorial doesn't make any sense.
"Part of the initial excitement in programming is easy to explain: just the fact that when you tell the computer to do something, it will do it. Unerringly. Forever. Without a complaint.
And that’s interesting in itself.
But blind obedience on its own, while initially fascinating, obviously does not make for a very likeable companion. What makes programming so engaging is that, while you can make the computer do what you want, you have to figure out how."[0]
I think somewhere between 2016 and 2026 the market realized that programmers _love_ writing tools for themselves and others, and it went full bore into catering to the Bike Shedding economy, and now AI is accelerating this to an absurd degree.
Me too, I love writing tools for myself and end up yak shaving all the time but why there's a tutorial for a machine that understand human language? Just type down your inner monologue and it will do it.
honestly, the biggest reason i deep dove on proper .claude stuff, was because im a cheap ass. I saw someone mention their agents/ that delegates to cheaper models, and figure that was a way I could reign in my own overall usage, and its been true so far. Im sure im one of the very few heavy claude code users that still stubbornly sits on the pro version. It won't be forever, if i land an important contract or job, I'll pretty quickly hop to max or whatever, but for my own usage right now, im getting by.
Sure, maybe this stuff isn't crazy relevant 2 years from now, but right now? Giving your agent a clean way to navigate and delegate tasks to keep that context window clean? its 100% vital.
I continue to find the non-stop claude spam fascinating. Gemini and ChatGPT have been very good for my needs, Claude not so much. Every week, if not every day, Claude spam is all over this site. But barely a peep about Gemini or ChatGPT coding capabilities.
https://www.reddit.com/r/ClaudeAI/comments/1s7mkn3/psa_claud...
It's like if cars didn't advertise MPG, but instead something that could change randomly.
(disclaimer: I work with the author)
1. Estimate free invoicing the requests and letting users figure it out after the fact. 2. Somehow estimating cost and telling users how much a request will cost.
We have 1, we want 2.
Except wait, I guess that actually makes sense, because it says Sonnet 1M is billed as extra usage... but also WTF, why is Sonnet 1M billed as extra usage? So Opus 1M is included in Max, but if you want the worse model with that much context, you have to pay extra? Why the heck would anyone do that?
The screen does also say "For other/previous model names, specify with --model", so I assume you can use that to get 200K Opus, but I'm very confused why Anthropic wouldn't include that in the list of options.
What a strange UX decision. I'm not personally annoyed, I just think it's bizarre.
And as others have said it's very easy to burn token usage on the $100/month plan. It's getting to the point where it's going to very much make sense to do model routing when using coding tooling.
I'm guessing Opus eats up usage much, much faster. I don't know what's going on, since a lot of people are hitting limits and I don't seem to be.
using it on the weekend gets you more use than during weekdays 9-5 in US eastern time.
I decided to give opencode go a try today. It's $5 for the first month. Didn't get much success with Kimi K2, overly chatty, built too complex solutions - burned 40% of my allocation and nothing worked. ¯\_(ツ)_/¯.
But Minimax m2.7. Wow, it feels just like Claude Opus 4.6. Really has serious chops in Rust.
Tomorrow/Wednesday will try a month of their $40 plan and see how it goes.
It's almost like an evolution of bobby tables.
According to the quiz, I am a beginner!
I got it as well.
I can only assume the quiz itself was vibe-coded and not tested. What an incredible time we live in.
I could see this being great for true beginners, but for them it might be nice to have even some more basics to start (how do I open the terminal, what is a command, etc).
Why isn't this just kinda seen as a repeat of the original birth of computers? Consider the IBM 350 (3.5mb) rented in the 50s for thousands per month. Now I have a drawer filled with SD cards that go up to 128gb that i cant even give away.
I'd be curious to know more about how this compares to other approaches.
The linked site is a pretty good interactive Claude tutorial for beginners.
And that’s interesting in itself.
But blind obedience on its own, while initially fascinating, obviously does not make for a very likeable companion. What makes programming so engaging is that, while you can make the computer do what you want, you have to figure out how."[0]
- [0] https://www.brynmawr.edu/inside/academic-information/departm...
I wouldn't have the thought to say to the machine to compact its context if I didn't know it has context and it can be compacted, right?
Sure, maybe this stuff isn't crazy relevant 2 years from now, but right now? Giving your agent a clean way to navigate and delegate tasks to keep that context window clean? its 100% vital.
edit: hop to max*