Models, catalog & pricing

Every model is addressed by a short, stable public id (the string you put in the model field). Public ids are attributed to the lab that built the model. They never leak an internal or upstream slug, and they don't change under you when routing or infrastructure changes behind them.

Browse the full catalog on the Models page, or list the ids programmatically with the endpoint below.

New: deepseek-v4.1-flash-1017, lightning fast

DeepSeek's October 17 update to V4.1 Flash is lightning fast, with leaner thinking that finishes sooner. Text and images in, tools, JSON mode, thinking you can switch on or off, and a 1M-token window. Launch price $0.13 in, $0.004 cached, $0.37 out per 1M tokens. On every coding plan, Taster included, and on the always-on Flash floor.

GET /v1/models

Returns the catalog in the standard OpenAI shape. No auth required to browse. This endpoint is public so a client can discover ids before a key is configured.

curl https://api.pgsgrove.com/v1/models

{
  "object": "list",
  "data": [
    { "id": "glm-5.2",        "object": "model", "created": 1700000000, "owned_by": "Z.ai" },
    { "id": "glm-5.2-turbo",  "object": "model", "created": 1700000000, "owned_by": "Z.ai" },
    { "id": "deepseek-v4-pro","object": "model", "created": 1700000000, "owned_by": "DeepSeek" },
    ...
  ]
}
  • owned_by is the lab that made the model.
  • Every turbo-capable model appears twice: once as its base id and once with a -turbo suffix (see Turbo). Both are valid model values.
  • The OpenAI SDKs expose this as client.models.list().

New models show up here the moment they go live, so this is the reliable way to keep an integration's model picker current.

Chat model prices

Every chat model on the API, at the rates you are billed. US dollars per 1M tokens. Cached input is applied automatically when a prompt prefix is reused (see Prompt caching); "n/a" means the model has no cached rate and reused input bills at the normal input rate. Turbo ids bill at 1.5x the standard rate for priority routing.

Public id Lab Context Image in Input Cached in Output Turbo
deepseek-v4-flash DeepSeek 1M no $0.12 $0.025 $0.23 yes
deepseek-v4-flash-0731 DeepSeek 1M no $0.12 $0.025 $0.23 yes
deepseek-v4.1-flash DeepSeek 1M yes $0.26 $0.01 $0.80 yes
deepseek-v4.1-flash-1017 DeepSeek 1M yes $0.13 $0.004 $0.37 yes
deepseek-v4-pro DeepSeek 1M no $1.65 $0.13 $3.30 yes
deepseek-v4-pro-0813 DeepSeek 1M no $1.65 $0.13 $3.30 yes
gemma-4-31b Google 256K yes $0.25 n/a $0.50 yes
ling-3.1-flash inclusionAI 256K no $0.38 $0.08 $1.15 no
minimax-m3 MiniMax 512K yes $0.38 $0.075 $1.50 no
kimi-k2.6 Moonshot AI 256K yes $0.95 $0.19 $4.40 yes
kimi-k3 Moonshot AI 1M yes $3.75 $0.45 $15.00 no
nemotron-3-super NVIDIA 256K no $0.11 n/a $0.50 yes
nemotron-3.5-super-vl NVIDIA 256K yes $0.13 $0.10 $0.50 no
nemotron-3-ultra NVIDIA 256K yes $0.65 $0.13 $2.80 yes
qwen-3.8-27b Qwen 256K no $0.52 $0.052 $3.90 no
mimo-v2.6-flash Xiaomi 1M yes $0.18 $0.004 $0.36 no
mimo-v2.6-pro Xiaomi 1M yes $0.56 $0.005 $1.12 no
glm-5.3-flash Z.ai 1M no $0.19 $0.04 $0.63 no
glm-5.2 Z.ai 1M no $0.65 $0.12 $2.00 yes
glm-5.3 Z.ai 1M no $1.60 $0.32 $5.20 no

Models without a turbo variant ignore a priority request and bill at the standard rate. See Turbo. Image-capable models are covered in Multimodal input.

Speech & embeddings

Two models are not chat models and are served on their own endpoints:

Public id Purpose Endpoint Price
kokoro-82m Text to speech POST /v1/audio/speech $1.25 per 1M input characters ($1.875 turbo)
embeddinggemma-300m 768-dim text embeddings POST /v1/embeddings $0.0025 per 1M input tokens

See Speech & embeddings. Sending one of these ids to /v1/chat/completions (or a chat id to the speech/embeddings endpoints) returns a clean 400 telling you which endpoint the model belongs to.