The short answer
Claude models are included in Whizi, and there is no Anthropic subscription to buy. Pro is the lowest paid plan that reaches Claude: it unlocks the Sonnet class and Claude Haiku. The Opus class and Claude Fable 5 require Powerhouse. Starter includes no Claude model at all.
| Plan | Price | Claude models included | Credits per month |
|---|---|---|---|
| Free | no charge | any text model, inside a lifetime allowance of 7 messages | 0 |
| Starter | $15.99/month, or $10.99/mo billed annually at $131.88 | none | 400 |
| Pro | $29.99/month, or $19.99/mo billed annually at $239.88 | Claude Sonnet 5, Sonnet 4.6, Sonnet 4.5, Haiku 4.5 | 2,000 |
| Powerhouse | $49.99/month, or $34.99/mo billed annually at $419.88 | the rest of the catalogue, including the Opus class and Claude Fable 5 | 8,000 |
A free account is not model gated. The chat path returns before the model gate for the free tier, so a prospect can send a Claude message inside the 7 message lifetime allowance. That allowance never resets.
Every Claude model, with credit cost and plan
A credit is one message on the house model, Whizi AI. A Claude message spends the number of credits on its own row, and nothing about message length, conversation length or answer length changes it.
The cost column below is what one standard answer costs to run at provider list rates. A standard answer is 1,000 input tokens plus 500 output tokens, and the rates were read from OpenRouter on 20 August 2026.
| Model | Plan | Credits per message | Model context window | Cost of one standard answer |
|---|---|---|---|---|
| Claude Haiku 4.5 | Pro | 4 | 200K | $0.0035 |
| Claude Sonnet 5 | Pro | 10 | 1M | $0.007 |
| Claude Sonnet 4.6 | Pro | 10 | 1M | $0.0105 |
| Claude Sonnet 4.5 | Pro | 10 | 1M | $0.0105 |
| Claude Opus 5 | Powerhouse | 20 | 1M | $0.0175 |
| Claude Opus 4.8 | Powerhouse | 20 | 1M | $0.0175 |
| Claude Fable 5 | Powerhouse | 50 | 1M | $0.035 |
Six further Claude rows carry a credit cost but are not in the priced comparison index, so there is no per-answer figure to publish for them.
| Model | Plan | Credits per message |
|---|---|---|
| Claude Sonnet 4 | Powerhouse | 10 |
| Claude Opus 4.7 | Powerhouse | 20 |
| Claude Opus 4.6 | Powerhouse | 20 |
| Claude Opus 4.5 | Powerhouse | 20 |
| Claude Opus 4.1 | Powerhouse | 80 |
| Claude Opus 4 | Powerhouse | 80 |
Every Claude model Whizi prices at 4 credits or more per message appears in those two tables. The 80 credit pair sits on the top rung in use, and that rung is Anthropic only: two Claude ids are priced there and nothing else in the catalogue is.
The Pro list names three Sonnet releases, so the oldest Sonnet resolves to Powerhouse despite carrying the same 10 credit rung as the three Pro can open.
What an allowance buys in Claude messages
Credits are one shared allowance, so these figures are what you get if you spend the whole period on a single Claude model and nothing else. The Haiku, Sonnet and current Opus arithmetic is already published in the credits reference. What follows is the part that page does not price out: the two dearest Claude rungs, and the weekly allowance.
| Plan and period | Credits | Claude Fable 5 at 50 | the 80 credit Opus rows |
|---|---|---|---|
| Starter, monthly | 400 | not included | not included |
| Pro, monthly | 2,000 | not included | not included |
| Powerhouse, monthly | 8,000 | 160 messages | 100 messages |
The monthly rows are the ones that decide a Claude budget today. Weekly billing is sold on mobile only, and the production value of the credit allowlist is the web client, so a weekly buyer is metered in messages rather than credits: 100 messages a week on Starter, 400 on Pro, 2,000 on Powerhouse. Each tier does set a weekly credit number explicitly rather than inheriting a fallback share of the monthly one, at Starter 100, Pro 500 and Powerhouse 2,000, but no weekly buyer is charged against those numbers while the allowlist reads web.
A 20 credit Opus turn on a balance of 3 is refused whole rather than part charged, because charging 3 for it would make the same model cost different amounts depending on when you sent it. A generation that fails or is aborted is refunded by removing its usage row. The rounding and flooring rules behind that are in the credits reference.
Using Claude without an Anthropic subscription
Whizi bills you, not Anthropic. There is no Anthropic account to connect and no Anthropic key to paste. Claude sits in the same model picker as the GPT, Gemini, Grok, DeepSeek, Llama, Mistral, Qwen and Kimi families, and the same credit allowance pays for all of them.
What that does not include is worth being straight about:
- No developer API. Whizi has no public API, so Claude in Whizi is a chat surface rather than something you can call from your own code.
- No team accounts. There is no seat-based or shared workspace plan.
- A Whizi turn budget, not the full window. A Claude turn is capped at 40,000 input tokens and 20,000 output tokens, so the 1M window on the rows above is a limit set by the model itself, not by what one Whizi turn can spend. Every Claude row priced in the index above carries a 200K window or larger, so none of them takes the reduced proportional budget that applies below a 93,000 token window.
- No Claude image, video or music generation. The image section is Whizi Image, Nano Banana, Flux and Stable Diffusion. Video is one row, Whizi Video AI, and audio is one row, Whizi Music AI. No Anthropic model appears in any of them.
- The agent personas do not run on Claude. No Anthropic model backs the Canadian CPA, study mentor, writing copilot or math tutor.
The Anthropic fast identifiers are excluded from the catalogue on purpose. They are the same base weights at a different latency tier for two to five times the money, so those rows carry no credit rung and cannot be selected at any tier.
If you are moving over, Claude is one of the two chat histories Whizi imports, alongside ChatGPT. The export is parsed in your browser. Thinking blocks, tool calls and attachments are dropped, so only the text of each turn is imported.
Switching into and out of Claude mid conversation
A conversation is not pinned to one model. Each turn is charged at the credit cost of whichever model answered it, so drafting on a 1 credit model and finishing on Claude costs 1 credit per draft turn and 10 or 20 for the answer you keep.
One behavior to know when you switch with a file in the thread: only the most recent user message carrying attachments has its attachments forwarded to the provider. If you attached a document six turns ago and then switch to Claude, re-attach it on the turn you want Claude to read it.
Auto is the other route into Claude. Auto has no credit cost of its own, because a turn costs whatever the model Auto picked costs. The rung Auto reaches for code is Claude Sonnet 5 at 10 credits, and difficulty is sticky inside a thread, so a follow-up to a code answer cannot drop back off that Claude row onto something cheaper. The published ranges are 1 to 1 on free and Starter, 1 to 10 on Pro, and 1 to 20 on Powerhouse.
More on the mechanics in switching models mid-conversation.
When Claude will not open
Three refusals cover the plan and allowance gates, and each carries an exact message.
Selecting a Claude model above your tier returns HTTP 403 with the code tier_upgrade_required and the message "Upgrade your plan to use this model." That is the Opus class on a Pro plan, or any Claude model at all on Starter. See model unavailable.
Running out of allowance returns HTTP 429 with the code message_limit_reached and the message "Your monthly message limit has been reached." Weekly plans say "weekly" instead. See out of credits.
A free account that has spent its 7 lifetime messages gets HTTP 402 with the code free_limit_reached and the message "Your free messages are used up. Start a subscription to keep chatting."
The gate is server side, and it is decided by the model identifier alone rather than by the conversation it sits in. That is why the oldest Sonnet refuses on Pro while the three Sonnet releases named on the Pro list open at the same 10 credit rung. How the resolution order works is covered in the models list.
How the Claude credit costs were set
A rung is derived from a reference turn of 3,000 input and 800 output tokens, divided by the cost of that same turn on the 1 credit anchor model. Models that fall between rungs round up, because the cheap direction is the one that costs money. The legal rungs are 1, 2, 3, 4, 5, 6, 8, 10, 12, 15, 20, 25, 30, 40, 50, 60, 80 and 100, and a model may not be priced anywhere else on the scale.
The Claude Sonnet class is one of the rows set by owner policy rather than arithmetic: it is charged 10 credits against a cost-true rung of 12.
One more thing the ladder guards against. A model identifier with no rung is charged 100 credits, the top of the scale, so an unpriced arrival surfaces as a support ticket rather than quietly running below cost. A build test fails if any catalogue model reaches production without a rung.
Credit pricing is enabled per client, and the production value of the allowlist is the web client. Off the allowlist the rungs on this page stop applying to what you are charged: an 80 credit Claude row costs one message against the older message allowance, exactly like a 1 credit row. The full scale across all families is in the credits reference, and the per-plan catalogue is in the models list.
- Pro is the lowest paid plan that reaches any Claude model
- The Sonnet class and Claude Haiku are the Pro set
- The Opus class and Claude Fable 5 require Powerhouse
- Claude Haiku is 4 credits, the Sonnet class is 10, Claude Opus 5 is 20, Claude Fable 5 is 50
- The 80 credit rung is the top rung in use, and only two Claude rows sit on it
- A free account can send a Claude message inside a lifetime allowance of 7 messages
- The flat turn budget is 40,000 input and 20,000 output tokens, and every Claude row in the price table carries a 200K window or larger
- A turn that does not fit the remaining balance is refused whole, not part charged
Frequently asked questions
Can I use Claude without an Anthropic subscription?
Yes. Pro at $29.99/month reaches the Sonnet class and Claude Haiku, and Powerhouse at $49.99/month adds the Opus class and Claude Fable 5. The one thing no Whizi plan reaches is the Anthropic fast latency tier: those identifiers carry no credit rung and are not selectable at any tier, so paying up does not put them in the picker.
How many credits does a Claude message cost in Whizi?
Claude Haiku is 4 credits, the Sonnet class is 10, the current Opus class is 20, Claude Fable 5 is 50, and the oldest Opus rows are 80. No Claude row sits at 5, 6, 8, 12, 15 or 25 credits. The rungs for every other family are in the credits reference.
Which Whizi plan do I need for Claude Opus?
Powerhouse, at $49.99/month or $34.99/mo billed annually at $419.88. Two more Claude rows sit above the Pro line with the Opus class: Claude Fable 5 at 50 credits, and the oldest Sonnet, which costs the same 10 credits as the Sonnet releases Pro can open but is not named on the Pro list.
Does Claude in Whizi have the full 1M context window?
Not in one turn. The number to hold on to is the turn budget: 40,000 input tokens and 20,000 output tokens, the flat allocation every catalogue model gets. One thing that surprises people comparing windows: OpenRouter counts input plus the requested maximum output against a window, not input alone, so the headline figure is not what a single request has free to spend.
Can I import my Claude chat history into Whizi?
Yes, within per request caps: 25 conversations at a time, and 400 messages per conversation, with the remainder sent as append slices onto the chat the first slice created. A raw .json export is accepted up to 300 MB, past which you are told to use the .zip export instead, which has no size ceiling of its own. Imports are keyed by conversation, so re-importing the same export or clicking Import twice lands the same chats once. If an import fails, see chat import errors.
What happens when a free account runs out of Claude messages?
Chat stops until you subscribe, and the refusal is the free_limit_reached one covered above. The part worth knowing before you register: a guest profile's usage rows are carried into the real account when the guest is claimed, so spending the allowance as a guest and then creating an account does not hand you a fresh one.