LLM API Pricing: Compare 62 Models Side by Side
Side-by-side API pricing for every major model — GPT-6 Astra, Sol and Luna, GPT-5.6, Claude, Gemini, Llama, Mistral, DeepSeek and more. Sort by input or output cost per token, see the context window, and find the cheapest model that does the job.
LLM API pricing comparison
LLM API pricing usually separates the tokens you send from the tokens a model generates. Input covers your prompt, instructions and supplied context; output covers the response. Rates below are in US dollars per million tokens. For example, 1M input tokens at $1 plus 250K output tokens at $4 costs $2 before tool fees or other charges. Compare both rates against your actual workload rather than choosing by one column alone.
Cached input can reduce the cost of repeated context, but only eligible cache hits receive that rate. Cache writes, storage, long prompts and processing modes may have separate charges. Open a model’s pricing page to check those conditions before estimating savings.
A cheaper model can also become more expensive per completed task if it needs longer answers, extra retries or escalation to another model. Test a representative set of requests and compare the total cost of successful results.
The shortlist below shows the 15 lowest listed output prices in this catalog, sorted by output cost first. Click a column header to reorder this shortlist. Use the full list of all 62 models to search providers, compare context windows and sort the entire catalog. Then use the cost calculator with your token mix and request volume.
Top 15 by lowest output price
Catalog list rates, before conditional discounts. Latest model check: 2026-11-11; review dates vary by model. The lowest token rate does not establish equal quality or availability.
| Model | Provider | Input /1M | Output /1M | Cost / call* | Context | Type |
|---|---|---|---|---|---|---|
| Qwen3 235B A22B Open weights, very cheap | Alibaba (Qwen) | $0.090 | $0.100 | $0.0002 | 256K | Fast / cheap open |
| Gemini 1.5 Flash-8B Cheapest capable model | $0.037 | $0.150 | $0.0002 | 1M | Fast / cheap | |
| Command R7B Cheapest Cohere tier | Cohere | $0.037 | $0.150 | $0.0002 | 128K | Fast / cheap |
| Llama 3.1 8B Open weights | Meta (Llama) | $0.200 | $0.200 | $0.0004 | 128K | Open weights open |
| MiMo v2.5 Open, cheap, 1M context | Xiaomi (MiMo) | $0.105 | $0.280 | $0.0004 | 1M | Fast / cheap open |
| Step 3.5 Flash Cheapest Step tier | StepFun | $0.090 | $0.300 | $0.0004 | 256K | Fast / cheap |
| Grok 3 mini | xAI (Grok) | $0.100 | $0.300 | $0.0004 | 128K | Fast / cheap |
| Llama 4 Scout Open weights (Groq) | Meta (Llama) | $0.110 | $0.340 | $0.0005 | 128K | Open weights open |
| Gemini 2.0 Flash | $0.100 | $0.400 | $0.0005 | 1M | Fast / cheap | |
| GPT-6 Luna Efficient model for focused, high-volume tasks · >272K input uses long-context rates | OpenAI | $0.100 | $0.500 | $0.0006 | 1.05M | Fast / cheap |
| Claude Haiku 5.5 Prompts up to 100K; above 100K, all token rates are 5× higher | Anthropic | $0.100 | $0.500 | $0.0006 | 1M | Fast / cheap |
| GLM-5.3-Flash 1M context; cached input $0.03/1M; cache storage temporarily free | Zhipu AI (GLM) | $0.150 | $0.500 | $0.0006 | 1M | Fast / cheap open |
| Grok 4 Fast 2M context | xAI (Grok) | $0.200 | $0.500 | $0.0007 | 2M | Fast / cheap |
| Mistral Small 4 | Mistral | $0.150 | $0.600 | $0.0007 | 128K | Fast / cheap |
| Command R | Cohere | $0.150 | $0.600 | $0.0007 | 128K | Balanced |
*Reference cost of one call with 1,000 input + 1,000 output tokens — a neutral yardstick. Use the calculator for your real usage. Lowest output-price row highlighted.
Model updates: Opus 5.5 pricing, Fast and Batch limits · Sol 6.1 pricing · Sol 6.1 vs Sol 6 and Astra · OpenAI usage tiers · Decisions API pricing
Go deeper
▣ Cost calculator
Enter your tokens and request volume to see the real monthly cost across every model, ranked cheapest first.
▤ Local VRAM calculator
Running models locally? Find out how much GPU VRAM a model needs at each quantization — and which card fits.
◇ Cheapest LLM API
The lowest-cost capable models for high-volume work, with the trade-offs that matter at scale.
◑ Image API pricing
Cost per image for GPT Image 2, Nano Banana & Nano Banana Pro (Gemini), FLUX, Ideogram and Seedream.
▶ Video API pricing
Native per-second list prices for MiniMax H3, Seedance 2.5, and Wan 3.0, including resolution and reference-video billing.
⋮⋮ Embeddings pricing
Cost per 1M tokens for OpenAI, Gemini, Cohere, Voyage and open embedding models — with dimensions and max input.
Pricing by provider
Best LLM for…
Frequently asked questions
How is LLM API pricing calculated?
Almost every LLM API charges per token, split into an input (prompt) price and a usually higher output (completion) price, quoted per 1,000,000 tokens. Your token bill is (input tokens × input rate + output tokens × output rate) / 1,000,000, before other charges. Use your provider’s tokenizer for an exact count. Calculate your token costs →
Which LLM API is cheapest?
The lowest-cost choice depends on input length, output length, caching and whether the model can complete your task. A low token price can cost more overall if it leads to repeated retries or escalation. See the cheapest LLM API guide →
What is the difference between input and output token pricing?
Input tokens are what you send (the prompt, system message, context, documents). Output tokens are what the model generates. Output is typically 2–5× more expensive than input, so capping max output length is the fastest way to cut cost.
Are these prices up to date?
The latest recorded model price check is 2026-11-11. This is not a full-catalog review date. Each model page links to its official source and shows its own verification date. The AI market moves fast — always confirm the current price before committing to volume.