Claude token costs are hard to predict
5 posts from 5 people5d activeFading, +0 this week
Posts per day
The posts behind it
...without going through one their client software or paying the API pricing. It is really annoying as those are two different product categories and I'd like to use the client of my choice with an LLM. OpenAI allows you to do that, so I have a proxy running on my computer which...
Claude used to be the best, until Feb 2026
...are inconsistent. Every other day Claude can’t access this tool or that tool. It’s frustrating and disrupts workflow. Wasteful data use. 6. Token spending is completely opaque. You have tokens one week, they're gone the next week with no explanation. No visibility into how they...
Old PRU Plan - Actually Unusable
So I still have the old PRU plan from before they migrated everyone to AIC. However this old PRU model is becoming useless very fast.. Like I understand how we won't be getting new models, fine. But:. Claude sonnet 4.6: gone sept 1st. Gemini 3.5 Flash: gone oct 2nd. GPT 5.4:...
Claude used to be the best, until Feb 2026
...are inconsistent. Every other day Claude can’t access this tool or that tool. It’s frustrating and disrupts workflow. Wasteful data use. 6. Token spending is completely opaque. You have tokens one week, they're gone the next week with no explanation. No visibility into how they...
...looks way more expensive from my quick calculations. Has anyone else calculated it? Is there any way to get Kimi K3 or Qwen 3.8 Max at a similar cost to what we're all paying by subscription for Claude or Codex? If not, I'd push back and say these Chinese models are actually...
Companies and products named
About this problem
Evidence
5 posts from 5 people in 3 places, about 1 a week over 65 days. Mostly on Hacker News comments, Claude App Store reviews, r/GithubCopilot. Tools named alongside: ChatGPT, Codex, Perplexity, GitHub.
Frustration Frustration 3 of 3· Seen on Hacker News, App Store reviews, Reddit
How it was grouped
Posts that state a pain and match the "llm-cost" rule about Claude. First post Sep 2, 2026, latest Nov 5, 2026. Corroborated: 3 or more posts from 2 or more people or places. Method
The same problem elsewhere
LLM token costs are hard to predict and control
15 posts from 15 people on GitHub issues, Hacker News, Reddit.
Anthropic token costs are hard to predict
6 posts from 6 people on Hacker News, GitHub issues, Reddit.
OpenAI token costs are hard to predict
5 posts from 4 people on GitHub issues, Hacker News, Reddit.
Gemini token costs are hard to predict
3 posts from 3 people on Reddit, GitHub issues.
Claude Code token costs are hard to predict
3 posts from 3 people on Reddit, Hacker News, GitHub issues.
GitHub token costs are hard to predict
2 posts from 2 people on Reddit, GitHub issues. Unverified.
LlamaIndex token costs are hard to predict
2 posts from 2 people on GitHub issues. Unverified.
GitHub Copilot token costs are hard to predict
2 posts from 2 people on Reddit. Unverified.
Other Claude problems
Trouble with Claude verification
Top post: "Verification", one of 2 posts from 2 people on App Store reviews. Unverified.
Trouble with Claude account
Top post: "Allow changing the email address on a Claude account without creating a new account", one of 9 posts from 9 people on App Store reviews, Reddit, Community forums.
Claude app gets 1 and 2 star reviews
Top post: "iOS, daytime naps from Apple Health are merged into the night and disappear", one of 185 posts from 183 people on App Store reviews, Reddit, GitHub issues.
Trouble with Claude desktop
Top post: "Claude Desktop (Code tab) auto-denies CLI-native SendMessage, breaking subagent resumption; the desktop", one of 24 posts from 23 people on GitHub issues, Reddit.
Claude rate limits and usage caps block work
21 posts from 21 people on Reddit, App Store reviews, GitHub issues.
Claude fails to return valid structured output
15 posts from 12 people on GitHub issues.
Trouble with Claude anthropic
Top post: "anthropic: treat Bedrock Claude model IDs like their canonical names in the mid-conversation system message", one of 13 posts from 13 people on GitHub issues, App Store reviews, Reddit.
Claude lock-in makes switching costly
12 posts from 12 people on GitHub issues, Hacker News, App Store reviews.
Claude runs out of context on real work
11 posts from 9 people on GitHub issues, Reddit, Hacker News.
Claude writes buggy code that is hard to review
11 posts from 11 people on Reddit, Hacker News, Mastodon.
Claude login, SSO or OAuth breaks
11 posts from 11 people on GitHub issues, Reddit, App Store reviews.
Trouble with Claude agent
Top post: "Autopilot with Assisted approvals denies tool calls without prompting since agent runtime 1.0.108", one of 11 posts from 11 people on GitHub issues, Reddit.
History