An AI model aggregation platform providing unified API access to multiple large language models with cost optimization features.

TokenPony (小马算力) is an AI model aggregation service that offers a unified API interface compatible with OpenAI and Claude specifications. The platform allows developers to access various large language models through a single endpoint while supporting load balancing and cost optimization.

Key models available include:

Deepseek-v3-0324: Enhanced reasoning capabilities for mathematical and coding tasks
qwen3-coder-480b: 450B parameter code generation model with multi-language support
kimi-k2-instruct-0905: Trillion-parameter MoE model for code generation and creative writing

The service provides metrics including average TTFT <500ms and pricing starting under ¥7 per million tokens. The platform serves over 60,000 developers with monthly token consumption exceeding 90 billion.

An AI model aggregation platform providing unified API access to multiple large language models with cost optimization features.

Key models available include:

Deepseek-v3-0324: Enhanced reasoning capabilities for mathematical and coding tasks
qwen3-coder-480b: 450B parameter code generation model with multi-language support
kimi-k2-instruct-0905: Trillion-parameter MoE model for code generation and creative writing

Model	Speed	Latency	Tests
deepseek-v3.1	57.51 t/s	0.30s	5
deepseek-v3.2-exp	39.89 t/s	0.58s	5
hunyuan-a13b-instruct	143.03 t/s	3.85s	5
kimi-k2-instruct-0905	13.20 t/s	1.09s	5
qwen3-8b	150.65 t/s	5.56s	5
qwen3-next-80b-a3b-instruct	164.04 t/s	0.45s	5

Time	Model	Speed	Latency
Nov 14, 10:21 AM	deepseek-v3.2-exp	39.89 t/s	0.58s
Nov 3, 12:17 PM	hunyuan-a13b-instruct	143.03 t/s	3.85s
Sep 23, 04:24 AM	kimi-k2-instruct-0905	13.20 t/s	1.09s
Sep 23, 04:20 AM	deepseek-v3.1	57.51 t/s	0.30s
Sep 23, 03:46 AM	qwen3-next-80b-a3b-instruct	164.04 t/s	0.45s
Sep 23, 03:44 AM	qwen3-8b	150.65 t/s	5.56s

Model	Speed	Latency	Tests
deepseek-v3.1	57.51 t/s	0.30s	5
deepseek-v3.2-exp	39.89 t/s	0.58s	5
hunyuan-a13b-instruct	143.03 t/s	3.85s	5
kimi-k2-instruct-0905	13.20 t/s	1.09s	5
qwen3-8b	150.65 t/s	5.56s	5
qwen3-next-80b-a3b-instruct	164.04 t/s	0.45s	5

Time	Model	Speed	Latency
Nov 14, 10:21 AM	deepseek-v3.2-exp	39.89 t/s	0.58s
Nov 3, 12:17 PM	hunyuan-a13b-instruct	143.03 t/s	3.85s
Sep 23, 04:24 AM	kimi-k2-instruct-0905	13.20 t/s	1.09s
Sep 23, 04:20 AM	deepseek-v3.1	57.51 t/s	0.30s
Sep 23, 03:46 AM	qwen3-next-80b-a3b-instruct	164.04 t/s	0.45s
Sep 23, 03:44 AM	qwen3-8b	150.65 t/s	5.56s

TokenPony

Supported Models

Recent Test Records

TokenPony

Supported Models

Recent Test Records

TokenPony

About TokenPony

Supported Models

Recent Test Records

TokenPony

About TokenPony

Supported Models

Recent Test Records