Xorin documentation

Accounts, models, billing, and privacy

The composer selects the Chat model used for the next request. Settings control authentication, the optional Strategy model, and diagnostic data.

Use Xorin credits

Credit rates

Credits are spent at the model provider's list price, with no Xorin markup on language-model usage. Each request is rounded up to the nearest cent, with a minimum of $0.01. Rates are in USD and current as of November 2026; they can change as providers update their prices.

Language models

Prices are per million tokens. Cached input, such as project context reused across turns in a conversation, is billed at a discounted rate for supported models. Longer conversations, larger projects, and deeper thinking use more tokens. The models offered are managed by Xorin and can change at any time, not only with a new release.

ModelTierInputOutput
Gemini 2.5 Flash LiteEconomy$0.10$0.40
Claude Haiku 5.5Economy$0.10$0.50
DeepSeek V4.1 FlashEconomy$0.14$0.42
GPT-4o MiniEconomy$0.15$0.60
GPT-5.6 LunaEconomy$0.20$1.20
Gemini 2.5 FlashEconomy$0.30$2.50
Kimi K2.5Economy$0.57$2.85
Kimi K2.6Economy$0.65$2.72
Gemini 3.8 FlashEconomy$0.75$3.75
GPT-6.1 TerraEconomy$1.00$5.00
Claude Haiku 4.5Economy$1.00$5.00
Gemini 2.5 ProStandard$1.25$10.00
GLM 5.3Standard$1.40$4.40
Claude Sonnet 5.5Standard$2.00$10.00
GPT-6.1 SolStandard$2.00$10.00
Gemini 4 ArgonStandard$2.00$10.00
Kimi K3.1 ProStandard$2.00$10.00
GPT-4oStandard$2.50$10.00
Claude Sonnet 4.5Standard$3.00$15.00
Kimi K3Standard$3.00$15.00
Claude Opus 5.5 (default)Premium$4.00$20.00
Claude Opus 4.6Premium$5.00$25.00

Asset generation

Generated assets have a fixed price per item. If a generation fails, its charge is refunded to your balance.

AssetPrice
Image: sprite, texture, sprite sheet, icon, or background$0.13
Image previews (batch of 3)$0.39
3D prop, FBX (default)$0.33
3D prop, GLB$0.26
3D humanoid, rigged$0.59
3D prop from a reference image, GLB / FBX$0.39 / $0.46
3D humanoid from a reference image, rigged$0.72
Audio: sound effect, voice line, or ambient loop$0.13
Audio: music$0.26

Bring Your Own Keys

BYOK sends compatible requests directly using keys saved through Unity Editor preferences. The Settings panel currently accepts keys for supported language-model and search providers.

Availability varies

Image, 3D, and audio generation require Xorin credits. Some models are available only when using Xorin credits. Available BYOK models depend on the provider keys you supply.

Chat and Strategy model selection

Select the Chat model beside the interaction mode in the composer. It handles normal requests and executes an approved plan. The selection is stored with the conversation.

Configure an optional Strategy model in Settings. It creates and revises plans in Plan mode and can help the Chat model with difficult decisions in any mode. Selecting None makes the Chat model handle planning and disables separate consultation. Strategy calls may add latency or cost.

Both model pickers group available options into Economy, Standard, and Premium tiers based on the model's input and output token prices. New installations default to Claude Opus 5.5 at Faster for Chat and Claude Opus 5.5 at Balanced for Strategy. Models that require Xorin credits are hidden when using BYOK.

When using Xorin credits, the list of models comes from the Xorin service and can change at any time. If a model you saved is no longer offered, Xorin switches to its replacement, or to the default model, and shows a one-time note. With BYOK, Xorin uses the model list built into the package.

Updating from an earlier version

Xorin 2.2.0 moves existing Xorin-credit users to the new default models once. The What's new note names the model you were using, and you can change back in the model picker or Settings.

Thinking level

Models that support it offer a thinking-level choice. The Chat choice appears beside the Chat model and belongs to the active conversation. The Strategy choice appears below the Strategy model in Settings and applies to planning and consultation.

Choice When to use it
FasterQuicker answers with lighter thinking; the default for Chat.
BalancedThe model's own default level; the default for Strategy.
DeeperFor the hardest problems: a step can take minutes, and it uses more credits. Not for big builds.

The menu marks the recommended choice for each model, and hovering shows the model's actual level. Thinking-level choices are available only when using Xorin credits. Deeper thinking can take longer and consume more credits; it does not change the model's authorization or Xorin's safety boundaries. When a step is still thinking after a minute, the status reads Thinking deeply · this can take a few minutes. This is normal, and the cost and time keep updating.

Request cost summaries

While a request runs, Xorin shows its progress and cost so far. After it completes, Xorin displays its total cost and LLM token count below the response. Hover over the summary for a receipt-style breakdown of applicable Chat, Strategy, and asset-prompt-refinement model usage, including token categories, calls, iterations, and subtotals.

Generated images, 3D previews and models, and audio appear as separate asset-generation line items when billing information is available. Asset line items distinguish confirmed billed charges from charges whose confirmation is still pending. Token totals cover LLM usage only, not image pixels, provider credits, or audio duration.

When using Xorin credits, the cost shown and your spend limit use the amount actually billed for each model call. If you press Stop, Xorin shows what the request cost up to that point.

Telemetry and transcripts

Setting What it controls
Send usage stats to improve Xorin Controls local recording of product events such as tool categories, model usage, failures, and keep/undo outcomes. When using Xorin credits, telemetry required by the Xorin service for analytics and billing attribution is sent independently of this toggle.
Chat transcripts Local transcript files containing message content, tool calls, parameters, and usage data.
Show detailed work activity Shows retained execution steps for successful turns. This is a presentation preference and does not change tool selection, permissions, telemetry, transcripts, or log verbosity.
Log level Local Xorin logs at Info, Debug, or Trace. Higher levels contain more sanitized diagnostic detail.

Telemetry events do not contain full chat transcripts, but some events can include diagnostic error messages or project asset paths. Local transcripts are stored under ~/.xorin/chat_history/<project_id>/.

When you use Xorin credits, the Xorin service keeps a short title for each chat (the first 80 characters of your first message), your 👍/👎 ratings, whether you kept or undid each change, and basic setup details such as your Xorin, Unity, and operating-system versions. It does not keep your full chats. Xorin does not upload its logs in the background; logs are sent only inside a task report you choose to send. The Xorin service may also process and retain request data under its applicable privacy and retention policy.

Task reports and consent

Selecting Report on a task opens a request-specific support form. Report submission requires a signed-in Xorin account and is separate from the Send usage stats to improve Xorin and Chat transcripts settings.

If you leave Include the chat unticked, the report contains only counts and technical details, with no message text. If you tick it, the report also includes the chat up to that request, what Xorin changed, any errors, workflow details, and up to 100 log lines. Secrets are hidden and images are removed before sending. A screenshot is sent only when you explicitly choose a file or capture the Scene or Game view and leave the preview attached.

Descriptions, consented task content, and screenshot content are retained for up to 30 days. The remaining diagnostic report and confirmation ID are retained for up to 180 days. Review the complete task-reporting guide before attaching project content.

Protect your keys