Token101 DocsSupported Models
All models currently available through the Token101 API, with pricing, capabilities, and selection guidance.
Overview
Token101 currently offers 20+ models from four providers — Anthropic, OpenAI, Google (Gemini), and Qwen — covering both text generation and image generation.
- Anthropic — Claude Fable 5, Claude Opus 4.8, Claude Sonnet 4.6, Claude Haiku 4.5, and more
- OpenAI — GPT 5.x series + Codex series + image generation (13 models)
- Google — Gemini 2.5 / 3 series (4 models)
- Qwen — Qwen Max, Qwen Plus (2 models)
The table below shows per-token pricing for each model (USD / 1M tokens), matching the values returned by the API.
Model Catalog
| Provider | Model ID | Input $/1M | Output $/1M | Prompt Caching |
|---|---|---|---|---|
| Anthropic | claude-fable-5 | $15.00 | $75.00 | Yes |
| claude-haiku-4-5-20251001 | $1.50 | $7.50 | Yes | |
| claude-opus-4-6 | $7.50 | $37.50 | Yes | |
| claude-opus-4-7 | $15.00 | $75.00 | Yes | |
| claude-opus-4-8 | $7.50 | $37.50 | Yes | |
| claude-sonnet-4-5-20250929 | $4.50 | $22.50 | Yes | |
| claude-sonnet-4-6 | $4.50 | $22.50 | Yes | |
| OpenAI | gpt-5 | $1.875 | $15.00 | Yes |
| gpt-5.1 | $1.875 | $15.00 | Yes | |
| gpt-5.1-codex | $1.875 | $15.00 | Yes | |
| gpt-5.1-codex-max | $1.875 | $15.00 | Yes | |
| gpt-5.1-codex-mini | $1.875 | $15.00 | Yes | |
| gpt-5.2 | $2.625 | $21.00 | Yes | |
| gpt-5.2-codex | $2.625 | $21.00 | Yes | |
| gpt-5.3-codex | $2.625 | $21.00 | Yes | |
| gpt-5.3-codex-spark | $2.625 | $21.00 | Yes | |
| gpt-5.4 | $3.75 | $22.50 | Yes | |
| gpt-image-1 | $7.50 | $60.00 | No | |
| Google Gemini | gemini-2.5-flash-lite | $0.15 | $0.60 | No |
| gemini-2.5-pro | $1.875 | $15.00 | No | |
| gemini-3-flash-preview | $0.75 | $4.50 | No | |
| Qwen | qwen-max | $3.60 | $14.40 | No |
| qwen-plus | $1.20 | $3.00 | No |
For a visual overview of all models and their capabilities, visit the models page.
Discovering Models
Token101 provides two model query endpoints:
| Endpoint | Auth | Description |
|---|---|---|
GET /api/v1/public/models | None required | Returns all models with pricing |
GET /api/v1/me/models | Login session (Dashboard) | Powers the Dashboard model list; adds an available field indicating whether each model is currently operational |
Every active model is available on every plan, including Free. The available field on the me/models Dashboard response only reflects temporary operational status (e.g. a model under maintenance), not plan restrictions.
For a version you can call with your API key, see Programmatic Model List below.
Programmatic Model List
Fetch the model list with your API key. GET /api/v1/models returns an OpenAI-compatible payload, so OpenAI SDKs and tools that already call /v1/models work without changes:
curl https://token.ppthub.shop/api/v1/models \
-H "Authorization: Bearer sk-xxxx...your-api-key"The response follows the OpenAI list shape. The token101 object is a Token101 extension carrying each model's service details:
{
"object": "list",
"data": [
{
"id": "claude-sonnet-4-5-20250929",
"object": "model",
"created": 0,
"owned_by": "anthropic",
"root": "claude-sonnet-4-5-20250929",
"permission": [],
"token101": {
"provider": "anthropic",
"display_name": "Claude Sonnet 4.5",
"status": "active",
"context_window_tokens": 200000
}
}
]
}Only models that are currently available (active and operational) appear in the list — a model under maintenance is omitted until it recovers. There is no available field to inspect here: absence from the list means the model is not usable right now.
Long-Context Pricing
Some models can apply a higher pricing tier once a request's token count crosses a large threshold (for example, 200,000 tokens). Among the models currently available, none charge a higher rate for long context: the large-context models (claude-opus-4-7, claude-sonnet-4-6, gpt-5.4, gpt-5.3-codex, gpt-5.2) use flat per-token pricing regardless of prompt length.
Per-token prices (in USD / 1M tokens) are shown on the models page and returned by GET /api/v1/models. If your client estimates cost before dispatching a large prompt, read the current price from there rather than assuming a fixed table.
For GPT-5.4, the billed input tokens include a fixed upstream service overhead on the order of ~4k tokens per request. Most of that overhead is charged at the discounted cache-read rate, so its cost impact is small — but do not assume the billed input equals exactly the tokens you sent.
Prompt Caching
Models that support prompt caching can cache repeated content, significantly reducing input cost on subsequent requests. Cache read discounts are typically 80–90% (i.e. cache hits are charged at only 10–20% of the input price).
Providers currently supporting prompt caching:
- Anthropic — All Claude models (Opus / Sonnet / Haiku)
- OpenAI — GPT 5.x series (
gpt-5.4,gpt-5.3-codex,gpt-5.2)
Writing new content into the cache is charged at a small premium (1.25× the input rate) — see Billing → Prompt Caching Costs for the complete pricing rules.
You can see how much caching has saved you in Usage Overview.
Reasoning Token Discounts
Some models generate reasoning tokens during inference. Token101 offers discounts on these tokens to reduce the cost of reasoning-intensive tasks:
| Model Series | Discount | Description |
|---|---|---|
Codex series (gpt-5.x-codex*) | 50% | Reasoning tokens charged at half the output price |
Non-Codex GPT series (gpt-5.x) | 25% | Reasoning tokens charged at 75% of the output price |
| Claude / Gemini / Qwen | No discount | Charged at standard output price |
Reasoning tokens are a subset of output tokens — they are not charged on top. The discount is applied directly to the output rate.
Choosing a Model
Everyday Coding & General Tasks
claude-sonnet-4-5-20250929— Balanced pick for coding and long-context assistance, with prompt caching supportgpt-5.2— OpenAI's general reasoning workhorse, also supports prompt caching
Complex Reasoning & Deep Analysis
claude-opus-4-8— Strongest reasoning capability for complex analysis that justifies the higher token costgpt-5.4— OpenAI's flagship reasoning model
Code-Specific
gpt-5.3-codex— Optimized for code generation, with 50% reasoning token discountclaude-opus-4-7— Strong coding model with a 1M-token context window for large repositories
Lightweight & Background Tasks
claude-haiku-4-5-20251001— Fast and lightweight, available on the free planqwen-plus— Low cost, suitable for classification, extraction, and background automation
Chinese Conversation
qwen-plus— Top choice for Chinese-heavy conversational flows at lower costqwen-max— High-quality Chinese output
Image Generation
Image generation is temporarily unavailable. Check the models page for the current status before integrating image endpoints.
Next Steps
- One-Line Setup — Configure Claude Code with a single command
- Billing — Learn about credits billing and top-up options
- Rate Limits — Understand request quotas and throttling rules
- Error Handling — Common error codes and troubleshooting