Cheapest LLM APIs, by capability band
The lowest standard-tier API prices for language models, ranked inside capability bands so a cheap small model is never presented as equivalent to a frontier model.
The cheapest LLM APIs by capability band (Artificial Analysis Intelligence Index): 50 and above: Claude Sonnet 5.5 at $2.00 in / $10.00 out per 1M tokens; 40 to 50: Ling 3.1 Flash at $0.075 in / $0.225 out per 1M tokens; 30 to 40: DeepSeek V4 Flash 0731 at $0.0598 in / $0.1795 out per 1M tokens; below 30: DeepSeek V4 Flash 0423 at $0.0181 in / $0.0361 out per 1M tokens. Cheapest overall: Mistral Nemo at $0.0178 blended.
Intelligence 50 and above
| Model | Input / 1M | Output / 1M | Blended 3:1 | Cheapest standard provider | Context (tokens) | AA intelligence | AA coding |
|---|---|---|---|---|---|---|---|
| Claude Sonnet 5.5Anthropic | $2.00 | $10.00 | $4.00 | Anthropic | 1,000,000 | 56.0 | Not reported |
| Gemini 4 ArgonGoogle DeepMind | $2.00 | $10.00 | $4.00 | Google AI Studio | 1,048,576 | 52.6 | Not reported |
| GPT-6.1 SolOpenAI | $2.00 | $10.00 | $4.00 | OpenAI | 1,050,000 | 51.8 | Not reported |
| Claude Opus 5.5Anthropic | $4.00 | $20.00 | $8.00 | Anthropic | 1,000,000 | 57.6 | Not reported |
| Claude Opus 5Anthropic | $5.00 | $25.00 | $10.00 | Anthropic | 1,000,000 | 50.8 | 78.0 |
| Claude Fable 5.1Anthropic | $10.00 | $50.00 | $20.00 | Anthropic | 1,000,000 | 53.4 | 81.6 |
| GPT-6 AstraOpenAI | $10.00 | $50.00 | $20.00 | OpenAI | 1,050,000 | 52.7 | 76.9 |
7 models in this band.
Intelligence 40 to 50
| Model | Input / 1M | Output / 1M | Blended 3:1 | Cheapest standard provider | Context (tokens) | AA intelligence | AA coding |
|---|---|---|---|---|---|---|---|
| Ling 3.1 FlashinclusionAI | $0.075 | $0.225 | $0.1125 | Novita AI | 262,144 | 41.1 | Not reported |
| GLM 5.3 FlashZ.ai (Zhipu) | $0.075 | $0.25 | $0.1188 | DeepInfrafp4 | 1,048,576 | 41.8 | 71.5 |
| Claude Haiku 5.5Anthropic | $0.10 | $0.50 | $0.20 | Anthropic | 1,000,000 | 43.4 | Not reported |
| MiMo-V2.6-ProXiaomi | $0.43 | $0.87 | $0.54 | DeepInfrafp8 | 1,050,000 | 46.3 | Not reported |
| GLM 5.3Z.ai (Zhipu) | $0.50 | $1.51 | $0.7513 | Sail Researchfp8 | 1,048,576 | 44.8 | 74.8 |
| Step 5 PreviewStepFun | $1.00 | $2.70 | $1.43 | StepFunfp8 | 1,048,576 | 43.7 | Not reported |
| Gemini 3.8 FlashGoogle DeepMind | $0.75 | $3.75 | $1.50 | Google AI Studio | 1,048,576 | 40.9 | 76.3 |
| Muse Spark 1.3Meta | $1.25 | $4.25 | $2.00 | Meta | 1,048,576 | 48.1 | 75.8 |
| Kimi K3Moonshot AI | $0.504 | $8.10 | $2.40 | Sail Researchfp4 | 1,048,576 | 43.6 | 76.2 |
| Grok 4.6xAI | $2.00 | $6.00 | $3.00 | xAI | 500,000 | 44.3 | 76.8 |
18 models in this band; the 10 cheapest are listed.
Intelligence 30 to 40
| Model | Input / 1M | Output / 1M | Blended 3:1 | Cheapest standard provider | Context (tokens) | AA intelligence | AA coding |
|---|---|---|---|---|---|---|---|
| DeepSeek V4 Flash 0731DeepSeek | $0.0598 | $0.1795 | $0.0898 | Baidufp8 | 1,048,576 | 34.3 | 69.1 |
| DeepSeek V4.1 FlashDeepSeek | $0.09 | $0.18 | $0.1125 | Decartfp4 | 1,048,576 | 39.5 | Not reported |
| MiMo-V2.6-FlashXiaomi | $0.10 | $0.28 | $0.145 | Darkbloomfp4 | 1,050,000 | 37.9 | Not reported |
| DeepSeek V4 Pro 0813DeepSeek | $0.1333 | $0.40 | $0.20 | Baidufp8 | 1,048,576 | 36.0 | 68.8 |
| GPT-6 LunaOpenAI | $0.10 | $0.50 | $0.20 | OpenAI | 1,050,000 | 38.1 | Not reported |
| DeepSeek V4 Pro 0423DeepSeek | $0.1775 | $0.3549 | $0.2218 | Baidufp8 | 1,048,576 | 30.4 | 59.4 |
| GLM 5.2Z.ai (Zhipu) | $0.189 | $0.594 | $0.2903 | Baidufp8 | 1,048,576 | 33.7 | 68.8 |
| DeepSeek V4 Flash Vision ExpDeepSeek | $0.2156 | $0.6468 | $0.3234 | DeepInfrafp8 | 1,048,576 | 34.8 | 65.0 |
| Qwen3.8 27BQwen (Alibaba) | $0.039 | $1.25 | $0.3418 | Chutesfp8 | 1,000,000 | 33.7 | 68.1 |
| GPT-5.6 LunaOpenAI | $0.20 | $1.20 | $0.45 | OpenAI | 1,050,000 | 37.3 | 71.4 |
25 models in this band; the 10 cheapest are listed.
Intelligence below 30
| Model | Input / 1M | Output / 1M | Blended 3:1 | Cheapest standard provider | Context (tokens) | AA intelligence | AA coding |
|---|---|---|---|---|---|---|---|
| DeepSeek V4 Flash 0423DeepSeek | $0.0181 | $0.0361 | $0.0226 | Baidufp8 | 1,048,576 | 24.4 | 52.0 |
| Solar Mini 4Upstage | $0.05 | $0.20 | $0.0875 | Upstage | 524,288 | 24.1 | Not reported |
| Ling 3.0 Flash VLinclusionAI | $0.06 | $0.18 | $0.09 | DeepInfrafp16 | 131,072 | 24.6 | 57.0 |
| Solar Pro 4Upstage | $0.09 | $0.36 | $0.1575 | Upstage | 524,288 | 28.2 | 52.7 |
| Nemotron 3.5 Super VLNVIDIA | $0.10 | $0.40 | $0.175 | Crusoebf16 | 1,048,576 | 26.5 | Not reported |
| Hy3 previewTencent | $0.18 | $0.60 | $0.285 | GMI Cloudbf16 | 262,144 | 25.3 | 58.8 |
| MiniMax M2.7MiniMax | $0.21 | $0.84 | $0.3675 | GMI Cloudfp8 | 204,800 | 22.8 | 52.6 |
| MiMo-V2.5-ProXiaomi | $0.3045 | $0.609 | $0.3806 | GMI Cloudbf16 | 1,050,000 | 26.0 | 60.2 |
| MiniMax M3MiniMax | $0.23 | $0.96 | $0.4125 | CoreWeavefp4 | 1,048,576 | 29.2 | 58.6 |
| MiniMax M2.5MiniMax | $0.27 | $0.95 | $0.44 | Venice AI | 204,800 | 22.8 | Not reported |
41 models in this band; the 10 cheapest are listed.
Not yet rated
| Model | Input / 1M | Output / 1M | Blended 3:1 | Cheapest standard provider | Context (tokens) | AA intelligence | AA coding |
|---|---|---|---|---|---|---|---|
| ExtractScaleDown | $0.04 | $0.00 | $0.03 | ScaleDown | 1,000,000 | Not reported | Not reported |
| SummarizeScaleDown | $0.04 | $0.00 | $0.03 | ScaleDown | 1,000,000 | Not reported | Not reported |
| Ling 3.0 FlashinclusionAI | $0.021 | $0.063 | $0.0315 | Novita AI | 262,144 | Not reported | Not reported |
| Granite 4.0 MicroIBM Granite | $0.017 | $0.112 | $0.0408 | Cloudflare | 131,000 | Not reported | Not reported |
| Nex-N2.5-MiniNex AGI | $0.025 | $0.10 | $0.0438 | Nex AGIbf16 | 262,144 | Not reported | Not reported |
| Nemotron 3.5 LightningNVIDIA | $0.025 | $0.12 | $0.0488 | Darkbloomint4 | 262,144 | Not reported | Not reported |
| Qwen3.7 FlashQwen (Alibaba) | $0.03 | $0.13 | $0.055 | Alibaba | 1,000,000 | Not reported | Not reported |
| Schematron V2 TurboInference.net | $0.03 | $0.15 | $0.06 | InferenceNetbf16 | 128,000 | Not reported | Not reported |
| Mercury 2.5Inception | $0.04 | $0.15 | $0.0675 | Inception | 260,000 | Not reported | Not reported |
| Laguna XS 2.1Poolside | $0.06 | $0.12 | $0.075 | Poolsidefp8 | 262,144 | Not reported | Not reported |
| Hy-MT2-1.8BTencent | $0.044 | $0.177 | $0.0773 | Tencentfp8 | 8,192 | Not reported | Not reported |
| Gemma 4 26B A4BGoogle DeepMind | $0.042 | $0.22 | $0.0865 | Darkbloom | 262,144 | Not reported | Not reported |
| Nemotron 3 Nano 30B A3BNVIDIA | $0.05 | $0.20 | $0.0875 | Novita AIfp4 | 262,144 | Not reported | Not reported |
| Qwen3.5-9BQwen (Alibaba) | $0.08 | $0.13 | $0.0925 | Darkbloomfp4 | 262,144 | Not reported | Not reported |
| Schematron V2 SmallInference.net | $0.05 | $0.23 | $0.095 | InferenceNetbf16 | 128,000 | Not reported | Not reported |
Models released in the last 12 months without an Artificial Analysis score, cheapest first. Top 15.
Cheapest host for the most widely served models
| Model | Hosts (standard tier) | Cheapest host | Input / 1M | Output / 1M | Most expensive host | Usage today |
|---|---|---|---|---|---|---|
| GLM 5.3 | 42 | Sail Researchfp8 | $0.50 | $1.51 | Alibaba$2.80 / $8.80 | 942.62B |
| DeepSeek V4.1 Flash | 32 | Decartfp4 | $0.09 | $0.18 | Fireworks$0.45 / $1.80 | 5.41T |
| GLM 5.3 Flash | 31 | DeepInfrafp4 | $0.075 | $0.25 | Reka$0.06 / $1.60 | 1.73T |
| GLM 5.2 | 28 | Baidufp8 | $0.189 | $0.594 | Wafer$0.19 / $10.00 | 140.87B |
| DeepSeek V4 Flash 0731 | 26 | Baidufp8 | $0.0598 | $0.1795 | Cloudflare$0.44 / $1.32 | 620.6B |
| gpt-oss-120b | 23 | Near AIfp4 | $0.03 | $0.15 | Cerebras$0.35 / $0.75 | 86.23B |
| DeepSeek V4 Pro 0813 | 23 | Baidufp8 | $0.1333 | $0.40 | Venice$1.65 / $4.95 | 161.08B |
| Kimi K3 | 20 | Sail Researchfp4 | $0.504 | $8.10 | Alibaba$3.45 / $17.25 | 203.41B |
Method
Prices are USD per 1M tokens at the standard service tier in the default region: the cheapest qualifying provider per model, with flex, batch, priority and fast endpoints and unusually low outliers never used as the headline. “Blended” weights input and output 3:1. Bands use the Artificial Analysis Intelligence Index reported through OpenRouter (CC BY 4.0, attribution required; not an SF·SIX measurement). Scope: language models listed on OpenRouter with an SF·SIX model page, priced across all of their providers. Per-token prices are not cost per task: a model that needs more tokens or retries can cost more overall. Free endpoints are listed under free inference.
Frequently asked questions
What is the cheapest LLM API right now?
By standard list price, Mistral Nemo at $0.0166 per 1M input tokens and $0.0216 per 1M output tokens, via IO.NET (fp16). The cheapest model is rarely the best value: compare within a capability band.
What is the cheapest LLM with an intelligence score of 50 and above?
Claude Sonnet 5.5, at $2.00 in and $10.00 out per 1M tokens (Artificial Analysis Intelligence Index 56.0).
What is the cheapest LLM with an intelligence score of 40 to 50?
Ling 3.1 Flash, at $0.075 in and $0.225 out per 1M tokens (Artificial Analysis Intelligence Index 41.1).
Why do prices for the same model differ between providers?
Hosts set their own prices and may serve quantized weights (for example fp8 or fp4), different context limits or regional endpoints. The table of hosts shows the quantization when the provider reports it.