EN RU 中文
$ cat /models/claude

Claude & Gemini API — OpenAI-compatible, up to 90% below official prices

Claude Sonnet 5, Claude Opus 4.8, Claude Fable 5 and Gemini Flash/Pro behind one OpenAI-compatible endpoint — point any OpenAI SDK at our base_url and it just works.

Streaming, tool calling, images and PDF attachments supported. One prepaid balance shared with image and video generation.

$ cat /pricing
Claude Sonnet 5 $0.360 / $1.800 -88% per 1M tokens
Claude Opus 4.8 $0.900 / $4.500 -82% per 1M tokens
Claude Fable 5 $3.600 / $18.000 -64% per 1M tokens
Gemini 3.1 Light $0.065 / $0.390 -74% per 1M tokens
Gemini 3.5 Flash $0.130 / $0.780 -91% per 1M tokens
Gemini 3.1 Pro $0.520 / $3.120 -74% per 1M tokens
GPT-5.6 Luna $0.225 / $1.350 -78% per 1M tokens
GPT-5.6 Terra $0.562 / $3.375 -78% per 1M tokens
GPT-5.6 Sol $1.125 / $6.750 -78% per 1M tokens
$ cat example.sh 
from openai import OpenAI

client = OpenAI(api_key="<token>", base_url="https://sosana.art/api")
resp = client.chat.completions.create(
    model="claude-sonnet-5",   # or claude-opus-4-8, claude-fable-5,
    messages=[                 # gemini-flash, gemini-pro
        {"role": "user", "content": "Hello!"},
    ],
)
print(resp.choices[0].message.content)
$ cat /features
  • up to 90% cheaper Claude Sonnet 5 at $0.36/$1.8 per 1M tokens vs $3/$15 official — live discounts in the table above
  • OpenAI-compatible drop-in replacement: swap base_url in any OpenAI SDK, LangChain or LiteLLM
  • tool calling & attachments client-side tools, streaming, images and PDFs via standard content parts
  • no accounts juggling Claude and Gemini under one token — no Anthropic or Google accounts needed
$ cat /faq
▸ How is this cheaper than the official API?

We aggregate traffic across providers and pass the volume discount on. You pay per token from a prepaid balance — up to 90% below list prices.

▸ Is it really OpenAI-compatible?

Yes — a standard Chat Completions endpoint. Set base_url to https://sosana.art/api in any OpenAI SDK and use your Sosana token.

▸ Which models are available?

Claude Sonnet 5, Claude Opus 4.8, Claude Fable 5, Gemini Flash and Gemini Pro. Gemini aliases always route to the newest model of the tier.

▸ Does streaming work?

Yes, standard SSE streaming, plus tool calling and usage reporting via stream_options.

▸ Can I attach files?

Yes — images via https URLs or base64 data URLs, PDFs and other files as base64 data URLs.

$ ls /models
/models/nano-banana · /models/veo · /models/claude
$ cd / · docs · playground