Best Cheap LLM APIs
Cheap LLM APIs are useful for chatbots, summarization, routing, classification, and high-volume workflows. Price alone is not enough: check output quality, context length, latency, provider reliability, and structured output support.
Quick Answer
Start with the cheapest models for simple or high-volume tasks, then move up to stronger models for coding, reasoning, and customer-facing answers. The table below uses the lowest available price in the current dataset, so always verify live provider pricing before committing production traffic.
Cheapest Models by Input Price
| Model | Platforms | Input /1M | Output /1M | Context | Best Use |
|---|---|---|---|---|---|
| inclusionAI: Ling-2.6-flash | openrouter | $0.010 | $0.030 | 256K | High-volume prompts |
| IBM: Granite 4.0 Micro | openrouter | $0.017 | $0.112 | 128K | High-volume prompts |
| Mistral: Mistral Nemo | openrouter | $0.019 | $0.030 | 128K | High-volume prompts |
| Nex AGI: Nex-N2-Mini | openrouter | $0.025 | $0.100 | 256K | High-volume prompts |
| OpenAI: GPT-5 Nano (batch) | openrouter | $0.025 | $0.200 | 391K | High-volume prompts |
| Meta: Llama 3.2 1B Instruct | openrouter | $0.027 | $0.201 | 128K | High-volume prompts |
| LiquidAI: LFM2-24B-A2B | openrouter | $0.030 | $0.120 | 125K | High-volume prompts |
| OpenAI: gpt-oss-20b | openrouter | $0.030 | $0.130 | 128K | High-volume prompts |
| Qwen: Qwen3.7 Flash | openrouter | $0.030 | $0.130 | 977K | High-volume prompts |
| Amazon: Nova Micro 1.0 | openrouter | $0.035 | $0.140 | 125K | High-volume prompts |
| OpenAI: gpt-oss-120b | openrouter | $0.037 | $0.170 | 128K | High-volume prompts |
| Cohere: Command R7B (12-2024) | openrouter | $0.037 | $0.150 | 125K | High-volume prompts |
Cheapest Models by Output Price
| Model | Platforms | Input /1M | Output /1M | Context | Best Use |
|---|---|---|---|---|---|
| inclusionAI: Ling-2.6-flash | openrouter | $0.010 | $0.030 | 256K | Low-cost generation |
| Llama Guard 3 8B | openrouter | $0.484 | $0.030 | 128K | Low-cost generation |
| Mistral: Mistral Nemo | openrouter | $0.019 | $0.030 | 128K | Low-cost generation |
| Meta: Llama 3 8B Instruct | openrouter | $0.040 | $0.040 | 8K | Low-cost generation |
| Sao10K: Llama 3 8B Lunaris | openrouter | $0.040 | $0.050 | 8K | Low-cost generation |
| MythoMax 13B | openrouter | $0.060 | $0.060 | 4K | Low-cost generation |
| Meta: Llama 3.1 8B Instruct | openrouter | $0.050 | $0.080 | 128K | Low-cost generation |
| Mistral: Mistral Small 3 | openrouter | $0.050 | $0.080 | 32K | Low-cost generation |
| Google: Gemma 3 4B | openrouter | $0.050 | $0.100 | 128K | Low-cost generation |
| IBM: Granite 4.1 8B | openrouter | $0.050 | $0.100 | 128K | Low-cost generation |
| Mistral: Ministral 3 3B 2512 | openrouter | $0.100 | $0.100 | 128K | Low-cost generation |
| Qwen: Qwen2.5 7B Instruct | openrouter | $0.040 | $0.100 | 128K | Low-cost generation |
Best Cheap Models for Coding Shortlist
| Model | Platforms | Input /1M | Output /1M | Context | Best Use |
|---|---|---|---|---|---|
| Arcee AI: Coder Large | openrouter | $0.500 | $0.800 | 32K | High-volume prompts |
| DeepSeek: DeepSeek V3 | openrouter | $0.200 | $0.800 | 128K | High-volume prompts |
| DeepSeek: DeepSeek V3 0324 | openrouter | $0.270 | $1.120 | 160K | High-volume prompts |
| DeepSeek: DeepSeek V3.1 | openrouter | $0.250 | $0.950 | 160K | High-volume prompts |
| DeepSeek: R1 0528 | openrouter | $0.500 | $2.150 | 160K | High-volume prompts |
| DeepSeek: R1 Distill Llama 70B | openrouter, bailian | $0.800 | $0.800 | 128K | High-volume prompts |
| DeepSeek: R1 Distill Qwen 32B | openrouter, bailian | $0.290 | $0.290 | 125K | High-volume prompts |
| DeepSeek: DeepSeek V3.1 Terminus | openrouter | $0.270 | $1.000 | 160K | High-volume prompts |
| DeepSeek: DeepSeek V3.2 Exp | openrouter | $0.270 | $0.410 | 160K | High-volume prompts |
| DeepSeek: R1 | openrouter, siliconflow, bailian | $0.000 | $0.000 | 160K | High-volume prompts |
| Kwaipilot: KAT-Coder-Pro V2 | openrouter | $0.300 | $1.200 | 250K | High-volume prompts |
| Mistral: Codestral 2508 | openrouter | $0.300 | $0.900 | 250K | High-volume prompts |
Best Cheap Models for Chatbots
Look for predictable output quality, low output price, enough context for conversation history, and stable provider uptime.
Best Cheap Models for Summarization
Prefer models with larger context windows and low input cost, because summarization often sends many tokens in and fewer tokens out.
What to Check Before Choosing a Cheap LLM
- Output quality for your exact task, not just benchmark scores.
- Context length, rate limits, latency, and provider reliability.
- Tool calling, JSON mode, structured output, and safety controls if your app needs them.
- Total cost across input, output, retries, caching, and fallback models.