// Models · API prices
Cheapest LLM APIs
Price per 1M tokens, lowest first – official APIs, inference platforms and OpenRouter, every row with its source.
573 paid text-model priceschecked 2026-10-22no unofficial resellers
01Cheapest official APIs
Vendor list prices, per 1M tokens. Blended = 3:1 input:output, in USD (1 USD = 6.7 CNY).
| # | Model | Vendor | Input | Output | Blended $ | Context | Source |
|---|---|---|---|---|---|---|---|
| 1 | Qwen3.7-Flash | Alibaba (Qwen) | ¥0.2 | ¥0.8 | 0.052 | 1M | official ↗2026-10-10 |
| 2 | Qwen-Turbo | Alibaba (Qwen) | ¥0.3 | ¥0.6 | 0.056 | — | official ↗2026-10-10 |
| 3 | GLM-4.6V-FlashX | Zhipu (Z.ai) | ¥0.15 | ¥1.5 | 0.073 | — | official ↗2026-10-10 |
| 4 | Qwen-Flash | Alibaba (Qwen) | ¥0.15 | ¥1.5 | 0.073 | — | official ↗2026-10-10 |
| 5 | Qwen3-VL-Flash | Alibaba (Qwen) | ¥0.15 | ¥1.5 | 0.073 | — | official ↗2026-10-10 |
| 6 | Doubao-Seed-2.0-mini | ByteDance (Doubao) | ¥0.2 | ¥2 | 0.097 | 262K | official ↗2026-10-10 |
| 7 | Qwen3.5-Flash | Alibaba (Qwen) | ¥0.2 | ¥2 | 0.097 | 1M | official ↗2026-10-10 |
| 8 | Ministral 3 3B | Mistral AI | $0.1 | $0.1 | 0.1 | 131K | official ↗2026-10-10 |
| 9 | Hunyuan-A13B | Tencent (Hunyuan) | ¥0.5 | ¥2 | 0.131 | 131K | official ↗2026-10-10 |
| 10 | Hy-MT2-Pro (translation) | Tencent (Hunyuan) | ¥0.5 | ¥2 | 0.131 | — | official ↗2026-10-10 |
| 11 | Qwen-Long | Alibaba (Qwen) | ¥0.5 | ¥2 | 0.131 | — | official ↗2026-10-10 |
| 12 | GPT-5 nano | OpenAI | $0.05 | $0.4 | 0.138 | 400K | official ↗2026-10-10 |
| 13 | Ministral 3 8B | Mistral AI | $0.15 | $0.15 | 0.15 | 262K | official ↗2026-10-10 |
| 14 | Step-3.5-Flash | StepFun (阶跃星辰) | ¥0.7 | ¥2.1 | 0.157 | 262K | official ↗2026-10-10 |
| 15 | Mercury 3 | Inception | $0.1 | $0.35 | 0.163 | 1M | official ↗2026-10-20 |
| 16 | Qwen-Plus | Alibaba (Qwen) | ¥0.8 | ¥2 | 0.164 | 1M | official ↗2026-10-10 |
| 17 | Qwen3.5-35B-A3B | Alibaba (Qwen) | ¥0.4 | ¥3.2 | 0.164 | 262K | official ↗2026-10-10 |
| 18 | Gemini 2.5 Flash-Lite | $0.1 | $0.4 | 0.175 | 1.05M | official ↗2026-10-10 | |
| 19 | GPT-4.1 nano | OpenAI | $0.1 | $0.4 | 0.175 | 1.05M | official ↗2026-10-10 |
| 20 | Qwen3.8-Flash | Alibaba (Qwen) | ¥0.8 | ¥2.7 | 0.19 | 1M | official ↗2026-10-10 |
02Cheapest by capability tier
Capability = Artificial Analysis Intelligence Index (where published). Any legitimate channel: official, platform or OpenRouter.
Frontier (AA index ≥ 45)
| Model | AA | Channel | Input | Output | Blended $ | Quant |
|---|---|---|---|---|---|---|
| DeepSeek V4.1 Flash 1017DeepSeek | 45.1 | DeepInfra via OpenRouter | $0.098 | $0.294 | 0.147 | fp8 |
| MiMo-V2.6-ProXiaomi (MiMo) | 46.3 | DeepInfra via OpenRouter | $0.387 | $0.783 | 0.486 | fp8 |
| Muse Spark 1.3Meta | 48.1 | OpenRouter via OpenRouter | $1.25 | $4.25 | 2 | — |
| Qwen3.8-MaxAlibaba (Qwen) | 45.4 | Official | ¥12 | ¥36 | 2.69 | — |
| Grok 4.7xAI | 46.4 | Official | $2 | $6 | 3 | — |
| Claude Sonnet 5.5Anthropic | 56 | Official | $2 | $10 | 4 | — |
| Gemini 4 ArgonGoogle | 52.6 | Official | $2 | $10 | 4 | — |
| GPT-6 SolOpenAI | 47.6 | Official | $2 | $10 | 4 | — |
Mainstream (AA index 30–45)
| Model | AA | Channel | Input | Output | Blended $ | Quant |
|---|---|---|---|---|---|---|
| DeepSeek-V4-Flash-0731DeepSeek | 34.3 | DeepInfra | $0.06 | $0.18 | 0.09 | — |
| GLM 5.3 FlashZhipu (Z.ai) | 41.8 | GMICloud via OpenRouter | $0.075 | $0.25 | 0.119 | fp8 |
| MiMo-V2.6-FlashXiaomi (MiMo) | 37.9 | DeepInfra via OpenRouter | $0.133 | $0.266 | 0.166 | fp8 |
| Claude Haiku 5.5Anthropic | 43.4 | Official | $0.1 | $0.5 | 0.2 | — |
| GPT-6 LunaOpenAI | 38.1 | Official | $0.1 | $0.5 | 0.2 | — |
| DeepSeek V4.1 FlashDeepSeek | 39.5 | DeepInfra via OpenRouter | $0.14 | $0.42 | 0.21 | fp8 |
| MiniMax-M3.1-FlashMiniMax | 42.7 | Official | ¥1 | ¥4 | 0.261 | — |
| Qwen3.8 27BAlibaba (Qwen) | 33.7 | Chutes via OpenRouter | $0.039 | $1.25 | 0.342 | fp8 |
Lightweight (AA index < 30)
| Model | AA | Channel | Input | Output | Blended $ | Quant |
|---|---|---|---|---|---|---|
| Llama 3.1 8B InstructMeta | 6.9 | DeepInfra via OpenRouter | $0.02 | $0.04 | 0.025 | fp8 |
| Ling 3.0 FlashAnt (inclusionAI) | 20.1 | Novita via OpenRouter | $0.021 | $0.063 | 0.032 | unknown |
| Qwen-TurboAlibaba (Qwen) | 6.4 | Official | ¥0.3 | ¥0.6 | 0.056 | — |
| gpt-oss-20bOpenAI | 9 | DeepInfra via OpenRouter | $0.03 | $0.14 | 0.058 | bf16 |
| Mistral Small 3Mistral AI | 6.7 | DeepInfra via OpenRouter | $0.05 | $0.08 | 0.058 | fp8 |
| Gemma 3 4BGoogle | 4.8 | DeepInfra via OpenRouter | $0.05 | $0.1 | 0.063 | bf16 |
| Mercury 2.5Inception | 12.3 | OpenRouter via OpenRouter | $0.04 | $0.15 | 0.068 | — |
| gpt-oss-120bOpenAI | 11.6 | DeepInfra via OpenRouter | $0.037 | $0.17 | 0.07 | bf16 |
03Same model, cheaper channel
Official price vs the lowest legitimate third-party channel we track (blended USD per 1M tokens).
04Cheapest across all channels
Includes inference platforms and OpenRouter listings; one row per model.
| # | Model | Channel | Input | Output | Blended $ |
|---|---|---|---|---|---|
| 1 | Llama 3.1 8B InstructMeta | DeepInfra via OpenRouter | $0.02 | $0.04 | 0.025 |
| 2 | Mistral NemoMistral AI | DeepInfra via OpenRouter | $0.029 | $0.03 | 0.029 |
| 3 | ExtractScaleDown | OpenRouter via OpenRouter | $0.04 | $0 | 0.03 |
| 4 | SummarizeScaleDown | OpenRouter via OpenRouter | $0.04 | $0 | 0.03 |
| 5 | Ling 3.0 FlashAnt (inclusionAI) | Novita via OpenRouter | $0.021 | $0.063 | 0.032 |
| 6 | Llama 3 8B LunarisSao10K | Parasail via OpenRouter | $0.04 | $0.05 | 0.043 |
| 7 | Qwen3.7-FlashAlibaba (Qwen) | Official | ¥0.2 | ¥0.8 | 0.052 |
| 8 | Qwen-TurboAlibaba (Qwen) | Official | ¥0.3 | ¥0.6 | 0.056 |
| 9 | gpt-oss-20bOpenAI | DeepInfra via OpenRouter | $0.03 | $0.14 | 0.058 |
| 10 | Mistral Small 3Mistral AI | DeepInfra via OpenRouter | $0.05 | $0.08 | 0.058 |
| 11 | Nova Micro 1.0Amazon | OpenRouter via OpenRouter | $0.035 | $0.14 | 0.061 |
| 12 | Gemma 3 4BGoogle | DeepInfra via OpenRouter | $0.05 | $0.1 | 0.063 |
| 13 | Command R7B (12-2024)Cohere | OpenRouter via OpenRouter | $0.037 | $0.15 | 0.066 |
| 14 | Mercury 2.5Inception | OpenRouter via OpenRouter | $0.04 | $0.15 | 0.068 |
| 15 | gpt-oss-120bOpenAI | DeepInfra via OpenRouter | $0.037 | $0.17 | 0.07 |
FAQ
What is the cheapest LLM API right now?
Among official prices we track, Qwen3.7-Flash (Alibaba (Qwen)) has the lowest blended price: $0.052 per 1M tokens (input ¥0.2, output ¥0.8; checked 2026-10-10). Cheap models are usually small – compare by capability tier below.
What does “blended price” mean?
Blended = (3 × input price + 1 × output price) ÷ 4 per 1M tokens, the common 3:1 input-to-output ratio. CNY prices are converted at 1 USD = 6.7 CNY for ranking only; tables show each source’s own currency.
Why is the same model cheaper on some platforms?
Open-weight models can be hosted by anyone, so inference platforms compete on price; some serve quantised weights (e.g. FP8) or have lower rate limits. The quantisation column shows what each platform reports.
Are cheap “API relay” resellers a good idea?
Unofficial resellers are not in this ranking. OpenAI’s and Anthropic’s terms forbid reselling or transferring API keys, and a relay sees every prompt you send. See our API relay guide for the risks and legitimate alternatives.
Full price table & cost calculator → · Free LLM APIs → · API relays & aggregators → · Price changes →