Model pricing

Input/output in USD per million tokens.

Moonshot AI

ModelModel IDContextFormal Engines
Kimi K3kimi-k31M$2.10/$10.50

Z.ai

ModelModel IDContextFormal Engines
GLM 4.5glm-4-5128K$0.420/$1.54
GLM 4.5 Airglm-4-5-air128K$0.090/$0.590
GLM 4.6glm-4-6198K$0.380/$1.54
GLM 4.7glm-4-7128K$0.280/$1.22
GLM 4.7 Flashglm-4-7-flash198K$0.040/$0.280
GLM 5glm-5198K$0.420/$1.34
GLM-5 Turboglm-5-turbo200K$0.840/$2.80
GLM 5.1glm-5-1200K$0.670/$2.12
GLM 5.2glm-5-21M$0.670/$2.12
GLM 5.3glm-5-31M$0.980/$3.08
GLM 5.3 Flashglm-5-3-flash1M$0.050/$0.170
GLM 5.3 AbliteratedFixed rate; no additional discounts. Cached input: $0.200/M tokens.glm-5.3-abliterated1M$1.60/$2.70
GLM 5.3 Flash AbliteratedFixed rate; no additional discounts. Cached input: $0.100/M tokens.glm-5.3-flash-abliterated1M$0.200/$0.750

DeepSeek

ModelModel IDContextFormal Engines
DeepSeek V3.2deepseek-v3-2128K$0.180/$0.280
DeepSeek V4 Flashdeepseek-v4-flash1M$0.060/$0.120
DeepSeek V4 Flash 0731deepseek-v4-flash-07311M$0.040/$0.120
DeepSeek V4.1 Flashdeepseek-v4-1-flash1M$0.210/$0.840
DeepSeek V4 Prodeepseek-v4-pro1M$0.730/$1.45
DeepSeek V4 Pro 0813deepseek-v4-pro-08131M$0.780/$2.34
DeepSeek V4 Flash Vision Expdeepseek-v4-flash-vision-exp1M$0.264/$0.792
DeepSeek V4.1 Flash AbliteratedFixed rate; no additional discounts. Cached input: $0.050/M tokens.deepseek-v4.1-flash-abliterated1M$0.200/$0.800

Uncensored

ModelModel IDContextFormal Engines
uncensoreduncensored128K$0.140/$0.630
uncensored-role-playuncensored-role-play128K$0.350/$1.40

Point any client at one base URL.

OpenAI clients

Cursor, Cline, Kilo Code, aider, LangChain, the OpenAI SDK.

OPENAI_BASE_URL=https://formalengines.com/inference/v1
OPENAI_API_KEY=fe_live_your_key

Claude Code, OpenClaw

Anything speaking the Anthropic Messages API.

ANTHROPIC_BASE_URL=https://formalengines.com/inference
ANTHROPIC_API_KEY=fe_live_your_key

Then use the model ID from the table above. Credit is prepaid and there is no minimum commitment; a key takes about a minute.

Get an API key

Account discounts may apply except to fixed-rate models. Charges truncate to whole microdollars per request.Back to inference