OpenRouter pricing comparison isn't about "which plan is cheaper" — it's about three things: which of 400+ models costs less, whether the platform marks up prices, and what top-ups cost. There's no monthly subscription, only prepaid Credits billed per token at pass-through rates with no markup, but top-ups carry a 5.5% fee ($0.80 minimum). The cheapest paid model, Qwen Flash, starts at $0.03/1M input tokens, while GPT-5.6 Sol costs $30/1M output — nearly a 50x spread. This article tallies entry and flagship prices across 15 providers, free-model limits, and the direct-API vs OpenRouter math. Prices below are as of 2026-08-13; check OpenRouter for live rates.
Billing model: no monthly plan, just prepaid Credits
OpenRouter has no monthly or annual subscription, only prepaid Credits. After you top up at settings/credits, the API and OpenRouter Chat share one USD balance billed per model at its token rate (source). A real price comparison has to look at three parts: model token rates, the top-up fee, and free-model quota.
| Billing item | Rule | Notes |
|---|---|---|
| Inference token rate | Matches provider list price | No markup |
| Top-up fee | 5.5% ($0.80 minimum) | 5% for crypto |
| Free-model quota | 50/day by default | 1000/day after $10 in credits |
| BYOK service fee | $25,000 list-price inference/month free ($200,000 Enterprise) | 5% beyond that |
| Logging discount | 1% usage discount for optional prompt/completion logging | Optional |
15-provider price comparison: 100x entry gap, nearly 50x flagship gap
The cheapest paid model on OpenRouter is Qwen Flash at $0.03/1M input, and the priciest flagship is GPT-5.6 Sol at $30/1M output — more than an order of magnitude apart. Here are the 15 core providers, all priced per 1M tokens (input/output) (source):
| Provider | Entry tier (input/output) | Flagship tier (input/output) |
|---|---|---|
| OpenAI | Luna $0.10/$0.60 | Sol $5/$30 |
| Anthropic | Haiku 4.5 $1/$5 | Opus 5 $5/$25 |
| Flash Lite $0.30/$2.50 | 3.6 Flash $1.5/$7.5 | |
| DeepSeek | Flash 0731 $0.08/$0.18 | Pro 0813 $0.435/$0.87 |
| xAI | Build $1/$2 | Grok 4.6 $2/$6 |
| Qwen | 3.7 Flash $0.03/$0.13 | 3.8 Max $2/$6 |
| Z.ai GLM | 4.7 Flash $0.06/$0.40 | 5.2 $0.49/$1.54 |
| MiniMax | M2.5 $0.22/$0.90 | M3 $0.30/$1.20 |
| Kimi | K2.5 $0.57/$2.85 | K3 $3/$15 |
| Xiaomi MiMo | V2.5 $0.14/$0.28 | V2.5-Pro $0.435/$0.87 |
| Meta | Scout $0.10/$0.30 | Maverick $0.20/$0.70 |
| Mistral | Small 4 $0.15/$0.60 | Medium 3.5 $1.5/$7.5 |
| Perplexity | Sonar $1/$1 | Pro Search $3/$15 |
| NVIDIA | Nano $0.05/$0.20 | Ultra $0.60/$3.60 |
| Tencent | Hy3 preview $0.06/$0.21 | Hy3 $0.132/$0.528 |
Two patterns stand out. Chinese providers keep entry prices very low — DeepSeek, Qwen, Z.ai GLM, MiMo, and Tencent Hunyuan all price input under $0.10/M, making them ideal for high volume. Flagships, meanwhile, get expensive — Claude Opus 5 costs $5/$25 and Kimi K3 $3/$15, 20-30x their own entry tiers.
How to pick: match the model to the job
Don't just chase the cheapest option — OpenRouter's real value is switching between 400+ models with one API key, so "which one to pick" matters more than "which is cheapest." Break it into three scenarios:
High volume / concurrency: prefer low-cost Chinese models
For large-scale inference, batch processing, and agent subtasks, DeepSeek Flash at $0.08/$0.18, Qwen 3.7 Flash at $0.03/$0.13, and Z.ai GLM 4.7 Flash at $0.06/$0.40 are outstanding value. They rank among the most-used models on OpenRouter and deliver far lower long-context cost than peer closed flagships.
Coding and agents: Claude, GPT, Grok, Kimi
For writing code and long-horizon agents, Claude (Opus 5 $5/$25, Sonnet 5 $2/$10) and OpenAI's GPT-5.3-Codex ($1.75/$14) are the defaults. Grok Build 0.1 is a $1/$2 coding agent, while Kimi K3 offers ~1M context at $3/$15 for ultra-long documents and UI/code generation.
Multimodal and search: Gemini, Perplexity, MiniMax
For image/audio/video input, Gemini 3.x starts at Flash Lite $0.30/$2.50. For web search and cited Q&A, Perplexity Sonar starts at $1/$1. For multimodal agents, look at MiniMax M3 at $0.30/$1.20.
Direct API vs OpenRouter: is the 5.5% fee worth it?
OpenRouter doesn't profit from token spread — it only charges the 5.5% top-up fee. If you switch between many models, that fee is cheaper than the hassle of direct accounts; if you stick to one model, direct is cheaper. Here's the math (source):
| Dimension | Direct vendor | OpenRouter |
|---|---|---|
| Token rate | Official price | Same as official, no markup |
| Platform fee | None | 5.5% top-up fee ($0.80 min) |
| Model coverage | Single vendor | 400+ models, one balance |
| Switching cost | Separate account and top-up per vendor | One key for all |
| Failover | None | Provider fallback |
The takeaway is clear: heavy single-model users should go direct to save that 5.5%; multi-model switchers, evaluators, and anyone needing provider fallback pay the fee for convenience and get their money's worth.
Caveats
- Credits expire: unused Credits may expire one year after purchase — don't top up too much at once.
- Refunds are limited: refunds within 24 hours on the Credits page, but platform fees are non-refundable and crypto is never refundable.
- Free models are capped: 50/day by default, 1000/day after $10 in credits — fine for testing, not production.
- BYOK has a service fee: your own provider keys get a monthly no-fee allowance measured by list-price inference cost ($25,000/month on pay-as-you-go, $200,000/month on Enterprise), then 5% beyond that.
- Rates change: all prices pass through openrouter.ai/models, so actual billing follows the live rate.