Claude Ultra alternatives (formerly Max 20x): Ultra vs an API plan
Claude Ultra is Anthropic's top consumer plan for Claude chat and Claude Code, renamed from Max 20x on October 30, 2026, and metered by a weekly limit once your account is verified. Standard Compute is the cost-optimized alternative for agent work: the router sends each step to the cheapest capable model, Claude on the hard steps and cheaper models on routine ones, inside one monthly budget that any agent can use.
TL;DR
Buy Ultra if you are a verified, Claude-only heavy user whose week fits the weekly limit. Move agent work to Standard Compute when you run several agents or models, need team budgets or Europe or US rules, or keep reaching the weekly limit.
What Claude Ultra offers
Ultra $200/mo (called Max 20x until October 30, 2026); Max $100/mo; Pro $20/mo. Verified Ultra accounts have no five-hour limit and get 10x Pro's usage every week, with a separate weekly limit for Fable; until Anthropic verifies an account from its account and payment history, a temporary limit applies (phone verification at checkout does not remove it). Max keeps a five-hour limit and a weekly limit. Ultra includes monthly Claude API credits. The plan login works only in Claude Code and Anthropic's apps. (Verified on claude.com/pricing and Anthropic's Ultra and Max plan articles, 2026-11-08.)
- For one heavy user, more Claude usage for the money than the same spend at per-token API rates
- No five-hour limit once verified, so long Claude Code sessions run through the day
- Claude chat, projects, Claude Code and Desktop in one subscription
- No API setup: sign in and work
Why people look for an alternative
- The weekly limit still ends the week early for parallel or always-on agents
- New and unverified accounts get a temporary limit, so the plan you pay for is not the plan you get on day one
- Anthropic models only, and the plan login is for Claude Code and Anthropic's apps, not other agents or your own scripts
- One account per person: no shared team budget or per-key spending caps
Claude Max vs Claude Ultra vs the API vs Standard Compute
Ultra removes the five-hour limit but not the weekly one. The API removes both and bills every token. Standard Compute bills one monthly budget and routes each step to the cheapest capable model.
| Option | What you pay | How the limit works | Agents and models |
|---|---|---|---|
| Claude Max | $100 a month | A five-hour limit and a weekly limit | Claude Code and Anthropic's apps; Claude only |
| Claude Ultra | $200 a month | Weekly limit only once verified; temporary limit before that | Claude Code and Anthropic's apps; Claude only |
| Claude API | Per token at Anthropic's rates | Rate limits by usage tier; no weekly cap, no fixed bill | Any agent with an Anthropic key; Claude only |
| Standard Compute | From $19, $19 compute budget | One monthly budget; requests stop at 100%, no overage | Claude Code, OpenCode, Cline, Kilo Code and more; the router picks the cheapest capable model per step, or pin a model ID |
Anthropic plans checked November 8, 2026 on Anthropic's plan articles (linked).
Which one fits your work
| Pick | When |
|---|---|
| Standard Compute | Mixed-model agent work: routine steps go to cheaper capable models, so the budget is not spent on Claude for every file read and test run |
| Agents beyond Claude Code: OpenCode, Cline, Kilo Code, Aider and your own scripts, on one key | |
| Smaller budgets: plans from $19 a month, when you would buy Ultra only to stop hitting Pro or Max limits | |
| Teams: company, team and key budgets, with opt-in Europe or US processing rules; we don't train on or store your prompts | |
| Claude Ultra | You are one heavy Claude Code user, your account is verified and the weekly limit covers your week: Ultra is the better deal |
| You want Claude chat, projects and the agent bundled in one subscription | |
| You want Claude models only, with first-party support from Anthropic |
Switching takes one config change
Standard Compute is OpenAI-compatible, so any tool or SDK that lets you set a custom base URL migrates in minutes:
Base URL = https://api.stdcmpt.com/v1 API key = your Standard Compute key Model = standardcompute
Setup guides for OpenClaw, Hermes, OpenCode, Cursor, Cline, Aider and more are on the integrations page.
FAQ
What is Claude Ultra?
Claude Ultra is the plan Anthropic called Max 20x until October 30, 2026. The price did not change. What changed is the limit: verified Ultra accounts no longer have a five-hour limit, only a weekly one. The plan in between is still called Max and keeps both limits.
Does Claude Ultra have a weekly limit?
Yes. Ultra has no five-hour limit once your account is verified, but it has a weekly limit across all models and a separate weekly limit for Fable. Anthropic verifies accounts automatically; in Settings > Usage, no Current session bar means you are verified.
Claude Ultra vs the API: which is cheaper?
For one person using Claude Code heavily inside the weekly limit, Ultra, because the subscription buys more Claude usage than the same money spent per token. The API, or Standard Compute, wins when the work runs in agents the plan login does not cover, runs past the weekly limit, or is mostly routine steps a cheaper model can do.
Can I use Standard Compute in Claude Code instead of Ultra?
Yes. Set ANTHROPIC_BASE_URL to https://api.stdcmpt.com, use your Standard Compute key and set ANTHROPIC_MODEL and every ANTHROPIC_DEFAULT_*_MODEL slot to catalog IDs. Usage then bills your Standard Compute plan, not your Claude subscription. Other agents use https://api.stdcmpt.com/v1. Many heavy users keep Ultra for their own sessions and put background agents on Standard Compute.
