Conclusion first: for the same mainstream model, the vendor's official price and our actual price can differ by an order of magnitude; but "0.1 off" is the deepest current seller tier, not a fixed price every request gets. This article puts representative models from Claude, GPT, Gemini, DeepSeek, Qwen, GLM, Kimi, and MiniMax on one table, official vs actual side by side.
Data as of 2026-11-01 18:50 (Beijing time). The platform currently indexes 352 model archives across 49 vendors; 190 models in stock, of which 180 have official and seller quotes. Prices move with seller listings; the latest values are on the live price table.
| Vendor | Representative model | Official (¥ / M tokens, in / out) | Our actual (deepest tier) | Discount | Sellers |
|---|---|---|---|---|---|
| Anthropic | Claude Sonnet 5 | in 13.39 / out 66.95 | in 0.134 / out 0.670 | 0.1 off | 7 |
| OpenAI | GPT-5.6 Sol | in 26.78 / out 133.90 | in 0.268 / out 1.34 | 0.1 off | 12 |
| Gemini 3.1 Pro | in 13.39 / out 80.34 | in 1.34 / out 8.03 | 1 off | 7 | |
| DeepSeek | DeepSeek V4.1 Flash | in 2.00 / out 8.00 | in 0.020 / out 0.080 | 0.1 off | 9 |
| Alibaba | Qwen3.8 Max | in 12.00 / out 36.00 | in 0.120 / out 0.360 | 0.1 off | 7 |
| Zhipu | GLM-5.3 | in 8.00 / out 28.01 | in 0.080 / out 0.280 | 0.1 off | 8 |
| Moonshot | Kimi K3 | in 20.00 / out 100.00 | in 0.200 / out 1.000 | 0.1 off | 7 |
| MiniMax | MiniMax M3 | in 2.10 / out 8.40 | in 0.021 / out 0.084 | 0.1 off | 7 |
Official prices from vendors' public pricing, USD converted at 1 USD = 6.695 CNY; our actual = official × (1 − seller discount), at the current deepest tier. The table shows each vendor's representative model, not their cheapest or strongest; the full list is on the corresponding brand page.
LONG AI isn't a single-upstream forwarding markup: sellers list their unused, already-bought model quota at their own price; the platform routes by quote, load, and health, and fails over automatically when a seller drops. The gap comes from idle-quota resale, bulk buying, and regional pricing - not from different model capability.
Look at Claude, GPT, Kimi, GLM first: Anthropic, OpenAI, Moonshot, Zhipu. Brand pages list every model, official price, our price, and current discount.
DeepSeek, Qwen, MiniMax's Flash/mini tiers usually suit high-frequency calls: DeepSeek, Alibaba, MiniMax. Don't look only at unit price - check context, output cap, and current seller count too.
Images bill per image and video per second; they can't be compared directly to text's per-million-tokens. See image category prices, video category prices, or check specific models on vendor pages like Google.
No. 0.1 off is the current deepest seller tier; actual requests bill by available tier, seller load, and routing result. For a more stable price, check the model page's per-tier quotes, or set a budget cap in an acceptable range.
Official is the vendor's public price; ours is what you actually pay after sellers supply at their own quotes. The platform doesn't change model capability - it only matches, schedules, and settles; different sellers of the same model may have different tiers.
Multi-seller models auto-failover by quote, load, and health; with a single seller, keep a same-family backup model in your code and degrade to it when the primary fails. See the integration docs.
Current public scope: 352 model archives / 49 vendors; 190 in stock; 180 with official and seller quotes. Archive count ≠ in-stock count: archives include some models with no current seller quote. Official prices at 1 USD = 6.695 CNY; our price at the current deepest tier. Numbers update daily; cite as "as of 2026-11-01."
Data sources: live price table, monthly report, model archive. To call directly, see the integration docs or sign up.
GPT-6.2 cuts list prices 20%, yet platform discounts run deeper: gpt-6-astra trades at under 1% of list. What this week's AI price war means for buyer
Prompt caching stores a repeated prompt prefix and bills cache reads at 5-12% of normal input. Real prices, hit conditions, and how to cut your API bi
Twelve Claude versions are on sale. The newest opus-5-5 costs $8.91 a month; the oldest opus-4-6 costs $49.50. Full table and a pick for each workload
gpt-5.6-terra and deepseek-v4.1-flash win on price and reliability; gemini-3.8-flash is fastest. Effective prices, measured speed, and a three-tier pi
Point your base URL at the platform and one key covers all 138 on-sale models across 51 vendors, in both OpenAI and Anthropic request formats. Copy-pa