Check current availability, input, output, credits, and proof first. If an online trial is available, test the model on the page; use Studio for saved history, assets, and longer iteration.
Example output
Example; generating model unverified · Best for inbox replies, support follow-up, and calm operational communication.
chat thread
A partner asks if we can guarantee every Prompt template will use fully licensed media before launch.
Thanks for asking. We are treating reviewed examples as a launch gate, not a cosmetic task.
The current plan is to keep draft assets separate, move final examples to approved URLs, and document any compatibility behavior that remains.
I cannot frame that as a blanket guarantee until the final audit passes.
Next step: I can share the current audit status and remaining replacement list.
Output
Acknowledgment / Direct answer / Useful detail / Boundary / Next step
Best for
Inbox triage and quick internal Q&AFirst-pass rewrites for emails, docs, and support repliesBudget-sensitive internal summaries and rewritesFast extraction from notes, transcripts, and pasted text
Supports
Chat
Online trial
Try Claude Haiku 4.5
Use Claude Haiku 4.5 for fast Claude chat, lightweight rewrites, quick summaries, and lower-cost daily assistant work.
Input setup
Prepare this run
Estimated credits: 2+
Chat controls
Choose when Rivya can search the web, how much reasoning it should use, and whether to include model thoughts when the selected model supports them.
Thoughts
Let the model default decide, or explicitly include or hide thought traces when the provider supports them.
A partner asks if we can guarantee every Prompt template will use fully licensed media before launch.
Thanks for asking. We are treating reviewed examples as a launch gate, not a cosmetic task.
The current plan is to keep draft assets separate, move final examples to approved URLs, and document any compatibility behavior that remains.
I cannot frame that as a blanket guarantee until the final audit passes.
Next step: I can share the current audit status and remaining replacement list.
Example; generating model unverified · Best for inbox replies, support follow-up, and calm operational communication.
chat thread
Original reply: We cannot refund this. Try again later. Policy does not allow it.
Revised reply: I cannot approve a refund from this message alone, but I can help review the failed task. Please send the task ID and the time it ran so we can check the logs.
Tone change: firm but helpful.
Removed risk: no unsupported blanket policy claim.
Remaining caveat: refund depends on task review.
Example; generating model unverified · Useful when a reply is technically correct but too abrupt for a customer.
chat thread
Macro: Sorry this happened. We always investigate failed generations and will make it right once we know what went wrong.
Revised macro: Thanks for sending this in. Please share the task ID and approximate time of the failed generation so we can review the logs.
Required placeholders: task ID, task time, account email if needed.
Policy boundary: do not promise a credit adjustment before review.
Agent note: use only when the customer reports a failed generation.
Example; generating model unverified · Keeps repeated support replies useful without expanding policy claims.
chat thread
Clause says vendor may change AI subprocessors with notice on the website only. What should we ask?
Risk: notice may be easy to miss and may not give the team enough time to object.
Business impact: privacy, procurement, and customer commitments could be affected.
Question 1: can changes trigger email notice to the account owner?
Question 2: is there an objection window for material subprocessor changes?
Question 3: can regulated customer data be excluded from new subprocessors by default?
Example; generating model unverified · Preview for Lightweight Legal Redline, focused on input context, structured reply, and actionable next steps.
chat thread
Objections: too many models, unclear credits, privacy questions, exports are hard to find, and team members need approvals.
Theme 1: decision overload around model choice.
Theme 2: cost anxiety around credits and usage predictability.
Theme 3: trust and governance concerns around privacy and approvals.
Recommended reply: lead with defaults, credit estimator, and workspace controls.
Product follow-up: improve model recommendation, expose export actions, and document approval flows.
Example; generating model unverified · Preview for Customer Objection Cluster, focused on input context, structured reply, and actionable next steps.
chat thread
Notes: onboarding templates hit the milestone, asset storage cleanup is still open, page title review has follow-ups, and usage questions continue.
Progress: template coverage reached the current target and proof depth improved.
Risk: media storage cleanup remains open before launch.
Decision needed: whether page title reviews should be fixed before launch or tracked as P2.
Customer signal: usage confusion is still creating support volume.
Next focus: storage validation, usage estimator copy, and targeted title cleanup.
Example; generating model unverified · Preview for Weekly Exec Status Synthesis, focused on input context, structured reply, and actionable next steps.
Decision fit
When this model is the right choice
Decision fit
Fit signals
Better suited for low-latency, high-frequency use
Much cheaper token pricing than Sonnet or Opus
Text-first Claude project in Rivya's current front end
Useful for lightweight analysis and rewrites
Fits cost-sensitive everyday collaboration
Best-fit tasks
Use it when the task looks like this
Inbox triage and quick internal Q&AFirst-pass rewrites for emails, docs, and support repliesBudget-sensitive internal summaries and rewritesFast extraction from notes, transcripts, and pasted textHigh-volume assistant work where latency matters
Model details
Inputs, output, and credits to confirm
Provider
Anthropic
Category
Chat
Capabilities
Chat
Credit model
Base credits / 1M tokens: input 55, output 285
Input path
Prompt only
Prompt setup
Up to 20,000 chars
Developer access
Available via API
Call Claude Haiku 4.5 from Public API v1 after checking the model fields, reference media rules, and credit behavior.
Use Claude Haiku 4.5 for fast, repeatable text work: inbox triage, internal Q&A, email rewrites, support replies, note extraction, and short summaries. It is a practical Claude option when response time and lower token cost matter more than the deepest analysis.
What is Claude Haiku 4.5's real scope on Rivya right now?
On Rivya, Claude Haiku 4.5 currently runs as a text-first chat project.
Which controls or inputs matter most once you're evaluating Claude Haiku 4.5 for real work?
Give it the source text, the audience, the action you want, and a clear output format. Haiku has few setup choices on this page, so a focused brief matters more than adjusting controls.
When is Claude Haiku 4.5 worth token-based spend?
Rivya currently bills 55 input credits and 285 output credits per million tokens. That tier fits high-volume short tasks where Haiku's speed and lower Claude-family cost can reduce the expense of repeated turns.
When should I choose Claude Haiku 4.5 over lighter or more expensive chat models?
Choose Haiku over Sonnet or Opus for faster, lower-cost routine work. Choose Sonnet when writing judgment or nuance matters more, and Opus when the analysis is difficult enough to justify the higher token tier.
What can I test first on this page?
Start here when the first response should test inbox triage, quick internal Q&A, first-pass rewrites, support replies, or budget-sensitive summaries on one Claude model. That gives you a contained first turn to check the prompt, direction, and response quality before moving into a longer project.
What can I set up on this page?
This page is set up as the public chat entry for Claude Haiku 4.5. That keeps the first result centered on inbox triage and quick internal Q&A instead of forcing the work into a longer thread too early.
What should I line up before I send the first turn?
Before the first turn, it matters more to line up the task frame, source material, and how careful the answer needs to be than to hunt for extra controls. On a text-first chat option like Claude Haiku 4.5, that opening setup usually decides whether the first result is strong enough for inbox triage and quick internal Q&A.
Do I need to sign in before I send here, and how do credits work?
The page is publicly visible, but the actual first send is still gated behind sign in. Rivya currently bills this by usage: 55 input credits and 285 output credits per million tokens. You can use this page to frame the question and choose the model first, then sign in when you are ready to start the live thread.
When should I stay on this page, and when should I move into the full Studio?
Stay on this page while the job is still a first conversation and you are testing whether Claude Haiku 4.5 can get traction on inbox triage and quick internal Q&A from the first turn. Open the full Studio once the work needs saved context, attached files, follow-up rounds, or a thread you know you will keep expanding.
Compare alternatives
Other models to consider next
Chat
Claude Fable 5
A text-first Claude chat model on Rivya for narrative reframing, audience-aware drafts, and structured editorial work.
Why consider it
Consider it when the credit note fits your next run: Base credits / 1M tokens: input 800, output 4,000.
