Get started

# Models and pricing

Each model is priced at the cheapest provider serving it plus the 10% markup. Billing is in whole millionths of a dollar, rounded up once per request. There is no subscription, seat fee or minimum spend.

**On this page**

- [Models](https://inference.dagon.so/docs/models#models)
- [Reading /v1/models](https://inference.dagon.so/docs/models#reading-v1-models)
- [How the charge is computed](https://inference.dagon.so/docs/models#how-the-charge-is-computed)
- [The hold](https://inference.dagon.so/docs/models#the-hold)
- [Output cap](https://inference.dagon.so/docs/models#output-cap)
- [Worked example](https://inference.dagon.so/docs/models#worked-example)
- [What each outcome costs](https://inference.dagon.so/docs/models#what-each-outcome-costs)

## Models

Models and their rates in US dollars per million tokens, 10% markup included

| Model | Providers | Input | Cached | Output | Context | Max output | Inputs | Reasoning | Refusals | ZDR |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Z.ai [GLM-5.3 Abliterated](https://inference.dagon.so/models/glm-5.3-abliterated) via Refuseless, Experiential Labs, Silk Compute `glm-5.3-abliterated` | Refuseless, Experiential Labs, Silk Compute, zero data retention | $1.10 | $0.165 | $2.20 | 1M | 66K | Text | Low–Max | Removed | Some |
| DeepSeek [DeepSeek V4.1 Flash Abliterated](https://inference.dagon.so/models/deepseek-v4.1-flash-abliterated) via Refuseless, Experiential Labs, Lutheos, Silk Compute `deepseek-v4.1-flash-abliterated` | Lutheos, zero data retention, Refuseless, Silk Compute, zero data retention, Experiential Labs | $0.11 | $0.022 | $0.44 | 1M | 66K | Text | Low–Max | Removed | Some |
| Z.ai [GLM-5.3 Flash Abliterated](https://inference.dagon.so/models/glm-5.3-flash-abliterated) via Refuseless, Experiential Labs, Silk Compute `glm-5.3-flash-abliterated` | Refuseless, Experiential Labs, Silk Compute, zero data retention | $0.132 | $0.0275 | $0.55 | 1M | 66K | Text | Low–Max | Removed | Some |
| abliteration.ai [Abliterated Large Flash](https://inference.dagon.so/models/abliterated-model-large-flash) via abliteration.ai `abliterated-model-large-flash` | abliteration.ai, zero data retention | $1.10 | $0.11 | $2.20 | 1M | 66K | Text, image | Low–Max Off | Removed | All |
| abliteration.ai [Abliterated Model](https://inference.dagon.so/models/abliterated-model) via abliteration.ai `abliterated-model` | abliteration.ai, zero data retention | $1.10 | $0.11 | $3.30 | 262K | 66K | Text, image | Min–Max Off | Removed | All |
| abliteration.ai [Abliterated Large v2](https://inference.dagon.so/models/abliterated-model-large-v2) via abliteration.ai `abliterated-model-large-v2` Based on GLM-5.3, abliterated and post-trained by abliteration.ai, so it can behave differently from GLM-5.3 Abliterated (glm-5.3-abliterated). | abliteration.ai, zero data retention | $3.30 | $0.33 | $5.50 | 1M | 66K | Text | Low–Max | Removed | All |
| Moonshot AI [Kimi K3.1 Pro Abliterated](https://inference.dagon.so/models/kimi-k3.1-pro-abliterated) via Silk Compute `kimi-k3.1-pro-abliterated` | Silk Compute, zero data retention | $5.50 | $0.55 | $16.50 | 1M | 66K | Text, image | Low–Max Off | Removed | All |
| DeepSeek [DeepSeek V4.1 Flash Fast](https://inference.dagon.so/models/deepseek-v4.1-flash-fast) via Refuseless `deepseek-v4.1-flash-fast` | Refuseless | $0.011 | $0.0022 | $0.088 | 1M | 66K | Text, image | Low–Max Off | Standard | Unknown |
| MiniMax [MiniMax M3.1 Flash](https://inference.dagon.so/models/minimax-m3.1-flash) via Refuseless `minimax-m3.1-flash` | Refuseless | $0.022 | $0.0022 | $0.088 | 1M | 66K | Text, image | Low–Max | Standard | Unknown |
| DeepSeek [DeepSeek V4.1 Flash](https://inference.dagon.so/models/deepseek-v4.1-flash) via Silk Compute `deepseek-v4.1-flash` | Silk Compute, zero data retention | $0.022 | $0.0044 | $0.176 | 1M | 66K | Text, image | Low–Max Off | Standard | All |
| Z.ai [GLM-5.3 Flash](https://inference.dagon.so/models/glm-5.3-flash) via Lutheos, Cheaper Inference `glm-5.3-flash` | Lutheos, zero data retention, Cheaper Inference, zero data retention | $0.088 | $0.0374 | $0.462 | 1M | 66K | Text, image | High | Standard | All |
| Moonshot AI [Kimi K3.1 Pro](https://inference.dagon.so/models/kimi-k3.1-pro) via Refuseless, Silk Compute `kimi-k3.1-pro` | Refuseless, Silk Compute, zero data retention | $0.275 | $0.165 | $2.75 | 1M | 66K | Text, image | Low–Max | Standard | Some |
| Tencent Hunyuan [Hy4](https://inference.dagon.so/models/hy4) via Silk Compute `hy4` | Silk Compute, zero data retention | $1.76 | $0.33 | $3.52 | 1M | 66K | Text, image | High Off | Standard | All |

USD per 1M tokens, markup included. Build snapshot at price version `2026-11-11.4`, as of 2026-11-11. Prices follow the providers and may have changed since: the live list is [`GET /v1/models`](https://inference.dagon.so/v1/models) (no key), also as [models.json](https://inference.dagon.so/models.json) and [models.md](https://inference.dagon.so/models.md). Model ids are exact; any other name is `404 model_not_found`.

New, removed and renamed models, refusal labels, providers and price moves are dated in the [model changelog](https://inference.dagon.so/changelog); follow its [Atom feed](https://inference.dagon.so/changelog.xml) or read [/changelog.json](https://inference.dagon.so/changelog.json).

`abliterated-model-large-v2`: Based on GLM-5.3, abliterated and post-trained by abliteration.ai, so it can behave differently from GLM-5.3 Abliterated (glm-5.3-abliterated).

Until 2026-11-07, `abliterated-model-large-v2` was also served by Refuseless, Experiential Labs and Silk Compute and priced at the cheapest of them. abliteration.ai now serves it alone, at abliteration.ai's price; their GLM-5.3 Abliterated is `glm-5.3-abliterated`.

`abliterated-model-large-flash`, `abliterated-model`, `kimi-k3.1-pro-abliterated`, `deepseek-v4.1-flash-fast`, `minimax-m3.1-flash`, `deepseek-v4.1-flash`, `glm-5.3-flash`, `kimi-k3.1-pro` and `hy4` accept image input; the rest are text only. Context windows go up to 1,048,576 tokens.

## Reading /v1/models

`GET /v1/models` returns `{"object": "list", "data": [...], "price_version", "prices_updated_at", "markup_bps"}`. Each entry has the model's id, its cheapest provider's rates with the markup in `pricing_per_mtok_micros`, `context_length`, `max_output_tokens`, `input_modalities`, `endpoints` (which of `/v1/chat/completions`, `/v1/responses` and `/v1/messages` it is served on) and every provider's offer. A model is not served on an endpoint missing from its list: `400 unsupported_endpoint`.

curl

```curl
curl https://inference.dagon.so/v1/models
```

## How the charge is computed

Cost is exact integer micro-USD: `ceil(Σ tokens × rate × (1 + markup))`, at the rates of the provider that served the request. Cache reads bill at the discounted cached-input rate; cache writes and uncached input at the standard input rate. A request that falls back to another provider is billed at that provider's rate plus the markup, and its receipt names that provider and the price version.

### The hold

Before forwarding, the gateway holds the request's maximum cost: an upper bound of the prompt (its JSON size in UTF-8 bytes, plus a template allowance per turn and for tools; 16,384 tokens per image part) plus the full output cap, times `n` on chat. The bound is roughly 2 to 4 times the real input cost for ordinary English, and everything beyond the charge is released when the request settles. A non-text part other than an image (a document, file or audio part) holds the whole context window.

### Output cap

Without a cap the gateway writes 16,384 tokens for you. The largest cap is 65,536 tokens or the model's own limit, whichever is lower; a larger one is `400 max_tokens_too_large`. A smaller cap means a smaller hold, so set it when your balance is tight.

## Worked example

A 1,200 character prompt to Abliterated Large Flash on abliteration.ai (`abliterated-model-large-flash`), answered with 412 input tokens (256 cached) and 238 output tokens:

| Step | Amount |
| --- | --- |
| Held before sending | $0.037545 |
| Charged from reported usage | $0.000724 (724 µ$) |
| Released back to the balance | $0.036821 |

Rates of the abliteration.ai route at price version 2026-11-02.2.

## What each outcome costs

| Outcome | Charge | Receipt |
| --- | --- | --- |
| Completes normally | Upstream-reported usage (`usage_source: "upstream"`) | Yes |
| Succeeds without a usage report | The full hold (`"reserved"`) | Yes |
| Client disconnects, or the stream breaks | Upstream usage if its final frame arrived; otherwise the prompt bound plus the output received (`"estimated"`) | Yes |
| Upstream error frame before any output | Zero | Yes, a zero receipt |
| Upstream answers 4xx or 5xx | Nothing; a 5xx, 429 or 402 first tries the next provider | No |
| Gateway stops mid-request | Nothing; holds are released at the next start | No |

The charge never exceeds the hold. If reported usage prices above it, you are charged the hold and the receipt carries `held_micros`.

Last updated 2026-11-11

[Previous **Quickstart**](https://inference.dagon.so/docs/quickstart) [Next **Authentication**](https://inference.dagon.so/docs/authentication)