Datasets:

uplinkswitch
/

llm-api-pricing

pretty_name: LLM API Pricing (Claude, GPT, Gemini, DeepSeek, Qwen, Kimi, GLM, Grok)
license: cc-by-4.0
language:
  - en
size_categories:
  - n<1K
tags:
  - llm
  - pricing
  - llm-pricing
  - api-pricing
  - token-pricing
  - claude
  - gpt
  - gemini
  - deepseek
  - qwen
  - openai-compatible
  - tabular
configs:
  - config_name: chat_models
    default: true
    data_files:
      - split: train
        path: data/chat_models.csv
  - config_name: media_models
    data_files:
      - split: train
        path: data/media_models.csv
  - config_name: chat_price_history
    data_files:
      - split: train
        path: data/chat_price_history.csv

LLM API Pricing

Per-token API prices for 91 chat models and per-image, per-second and per-generation prices for 94 image, video, audio and 3D models. Each row puts the model developer's own list price next to the price on UplinkSwitch, an OpenAI- and Anthropic-compatible API gateway.

Every value comes from the public catalog API at https://api.uplinkswitch.com/v1/catalog and /v1/catalog/media. It's the same feed the gateway bills from, and it needs no key. The snapshot_at column shows when each row was read. Every row links to the model's price page on uplinkswitch.ai (model_url), and open-weight models also link to their official weights on the Hub (weights_repo).

Live, sortable version with a monthly cost calculator: LLM API Price Comparison Space
Official weights for every open model here: LLM API pricing: open-weight models
Same prices on the web: uplinkswitch.ai/models · price index (CSV/JSON) · side-by-side comparisons

Configs

Config	Rows	What one row is
`chat_models` (default)	91	One chat model: list and UplinkSwitch prices per 1M input, cached-input and output tokens
`media_models`	256	One price variant of an image, video, audio or 3D model (for example `resolution=2K`)
`chat_price_history`	258	One chat model on one snapshot date (2026-10-16 onward)

The raw catalog responses are in raw/catalog.json and raw/catalog_media.json.

Load it

from datasets import load_dataset

chat = load_dataset("uplinkswitch/llm-api-pricing", "chat_models", split="train").to_pandas()
media = load_dataset("uplinkswitch/llm-api-pricing", "media_models", split="train").to_pandas()
history = load_dataset("uplinkswitch/llm-api-pricing", "chat_price_history", split="train").to_pandas()

# Monthly cost of 50M input + 10M output tokens, list vs UplinkSwitch
chat["list_monthly"] = 50 * chat.list_input_usd_per_1m + 10 * chat.list_output_usd_per_1m
chat["uplink_monthly"] = 50 * chat.price_input_usd_per_1m + 10 * chat.price_output_usd_per_1m
print(chat.sort_values("uplink_monthly")[["display_name", "list_monthly", "uplink_monthly"]].head(20))

With pandas only:

import pandas as pd
chat = pd.read_csv("hf://datasets/uplinkswitch/llm-api-pricing/data/chat_models.csv")

Columns

`chat_models`

Column	Type	Meaning
`model_id`	string	The `model` value to send to the API
`display_name`	string	Model name as the developer writes it
`developer`	string	Anthropic, OpenAI, Google, xAI, DeepSeek, Qwen, Moonshot AI, Z.ai, MiniMax, Xiaomi, Tencent, NVIDIA, Meta, …
`wire_formats`	string	API formats the model accepts, pipe-separated (`openai`, `anthropic`)
`list_input_usd_per_1m`	float	Developer list price, USD per 1M input tokens. Empty when the developer publishes no price
`list_cache_read_usd_per_1m`	float	Developer list price for cached input tokens
`list_output_usd_per_1m`	float	Developer list price, USD per 1M output tokens
`price_input_usd_per_1m`	float	UplinkSwitch price, USD per 1M input tokens
`price_cache_read_usd_per_1m`	float	UplinkSwitch price for cached input tokens
`price_output_usd_per_1m`	float	UplinkSwitch price, USD per 1M output tokens
`pct_under_list_input`	float	`100 × (1 − price / list)` on input, rounded to 0.1
`pct_under_list_output`	float	Same, on output
`model_url`	string	The model's page on uplinkswitch.ai, e.g. `https://uplinkswitch.ai/models/claude-opus-5-5`
`weights_repo`	string	Official open weights on the Hugging Face Hub; empty for closed models
`snapshot_at`	string	ISO 8601 time the catalog was read

`media_models`

Column	Type	Meaning
`model_id`, `display_name`, `developer`	string	As above (OpenAI, Alibaba, ByteDance, Google, Kling, Tripo, Black Forest Labs, Independent, ElevenLabs, MiniMax, xAI)
`kind`	string	`image`, `video`, `audio` or `3d`
`unit`	string	What one price buys: `image`, `second`, `minute`, `generation` or `1k_characters`
`variant`	string	The parameter values this price applies to, e.g. `resolution=4K` or `duration=8, audio=True`; `default` when there is one price
`price_usd`	float	UplinkSwitch price per unit
`list_usd`	float	Developer list price per unit, empty when none is published
`pct_under_list`	float	`100 × (1 − price / list)`
`endpoints`	string	API paths that serve the model, pipe-separated
`model_url`	string	The model's page on uplinkswitch.ai, e.g. `https://uplinkswitch.ai/models/media/nano-banana-2`
`snapshot_at`	string	ISO 8601 time the catalog was read

`chat_price_history`

snapshot_date, model_id, display_name, list_input_usd_per_1m, list_output_usd_per_1m, price_input_usd_per_1m, price_output_usd_per_1m. Use it to track list-price and gateway-price changes over time.

Example rows (2026-10-19)

Model	List in / out ($/1M)	UplinkSwitch in / out ($/1M)	Family pricing
Claude Opus 5.5	4 / 20	2.10 / 6.00	All Claude models
GPT-6.1 Sol	2 / 10	0.60 / 3.00	All GPT models
Gemini 3.1 Pro Preview	2 / 12	1.20 / 7.20	All Gemini models
DeepSeek V4 Pro	1.32 / 3.96	0.792 / 1.92	All DeepSeek models
Kimi K3	3 / 15	0.90 / 4.50	All Kimi models
Qwen3.7 Max	1.475 / 4.425	0.885 / 2.655	All Qwen models
GLM-5.3	1.092 / 3.432	0.3276 / 1.0296	All GLM models
Grok 4.7	2 / 6	0.60 / 1.80	All Grok models

In this snapshot, 76 of the 91 chat models have a published list price. The median UplinkSwitch output price is 58.5% under list. Cached input costs more than list on five models (deepseek-v4-flash-0731, deepseek-v4-pro-0813, deepseek-v4.1-flash, gpt-6-luna, mimo-v2.6-pro), so check the cache column if your workload is cache-heavy.

Open-weight models: weights and API price

Model	Weights on the Hub	API price page
DeepSeek V4 Pro	deepseek-ai/DeepSeek-V4-Pro	deepseek-v4-pro
DeepSeek V4 Flash	deepseek-ai/DeepSeek-V4-Flash	deepseek-v4-flash
DeepSeek V4 Flash 0731	deepseek-ai/DeepSeek-V4-Flash-0731	deepseek-v4-flash-0731
DeepSeek V4 Flash Vision (exp)	deepseek-ai/DeepSeek-V4-Flash-Vision-Exp	deepseek-v4-flash-vision-exp
DeepSeek V4.1 Flash	deepseek-ai/DeepSeek-V4.1-Flash	deepseek-v4.1-flash
DeepSeek V4.1 Flash 1017	deepseek-ai/DeepSeek-V4.1-Flash-1017	deepseek-v4.1-flash-1017
GLM-5.3	zai-org/GLM-5.3	glm-5.3
GLM-5.3 Flash	zai-org/GLM-5.3-Flash	glm-5.3-flash
GLM-5.2	zai-org/GLM-5.2	glm-5.2
GLM-5.1	zai-org/GLM-5.1	glm-5.1
Kimi K3	moonshotai/Kimi-K3	kimi-k3
Kimi K2.6	moonshotai/Kimi-K2.6	kimi-k2.6
Kimi K2.5	moonshotai/Kimi-K2.5	kimi-k2.5
Qwen3.6 27B	Qwen/Qwen3.6-27B	qwen3.6-27b
MiniMax M2.7	MiniMaxAI/MiniMax-M2.7	minimax-m2.7
MiniMax M2.5	MiniMaxAI/MiniMax-M2.5	minimax-m2.5
MiMo V2.5	XiaomiMiMo/MiMo-V2.5	mimo-v2.5
DiffusionGemma 26B A4B	google/diffusiongemma-26B-A4B-it	diffusiongemma-26b-a4b-it
Llama 3.2 11B Vision	meta-llama/Llama-3.2-11B-Vision-Instruct	llama-3.2-11b-vision-instruct

Pricing write-ups that use this data, one per model family:

How it's built

scripts/build.py reads the two catalog endpoints and writes every file in data/ and raw/. It has no dependencies beyond the Python standard library. Run python scripts/build.py <out_dir> to rebuild it yourself. List prices are the model developers' own published per-token prices; the catalog stores them next to the gateway price so the two can be compared row by row.

Limitations

Prices change. Trust snapshot_at, and for a live figure read the catalog API or uplinkswitch.ai/models.
List prices are standard-tier, non-batch prices. Batch, priority, long-context and regional tiers aren't included.
Open-weight fine-tunes and community models often have no developer list price, so their list columns are empty.
Volume discounts (3% from 100M tokens, 6% from 1B, 10% from 10B per month) aren't applied in the price columns.

Use the prices

The same key and base URL work with the OpenAI and Anthropic SDKs:

from openai import OpenAI

client = OpenAI(base_url="https://api.uplinkswitch.com/v1", api_key="sk-uplink-live-...")
reply = client.chat.completions.create(model="claude-opus-5-5", messages=[{"role": "user", "content": "hi"}])

Python package: pip install uplinkswitch
SDK setup (OpenAI, Anthropic, Gemini): uplinkswitch.ai/guides/sdks
Coding tools: Claude Code · Cursor · Cline · Roo Code · Kilo Code · OpenCode
Daily token-count check of every model: uplinkswitch.ai/proof
Verify a signed usage statement: uplinkswitch.ai/statements/verify
Savings calculator: uplinkswitch.ai/savings
Get a key (prepaid from $10, no subscription): uplinkswitch.ai/console
Source and issues: github.com/uplinkswitch

License and citation

CC BY 4.0. Please credit "UplinkSwitch LLM API Pricing dataset" with a link to this page.

@misc{uplinkswitch_llm_api_pricing,
  title  = {LLM API Pricing},
  author = {{Uplink Switch LLC}},
  year   = {2026},
  url    = {https://huggingface.co/datasets/uplinkswitch/llm-api-pricing}
}