Pricing · 2026-07-29
LLM pricing comparison
List price per million tokens for every tracked model — next to the share of real gateway traffic it carries. Cheap only counts if the model can hold production load.
What LLM APIs cost in July 2026
The cheapest model with measurable traffic is inclusionAI: Ling-2.6-flash at $0.03 per million output tokens. Across all tracked models the gateway spends roughly $22.8M per day at list rates, so the gap between the cheapest and most expensive tiers is the single largest lever on an inference bill.
Output tokens dominate cost for most workloads: a typical chat or agent turn reads a short prompt and writes a long answer. Sort by output price first, then check whether the model appears in the usage column at all — an unused cheap model is untested, not a bargain.
Cheapest models actually in production
Under $1 per million output tokens, holding at least 0.1% of gateway traffic.
| Model | Lab | $/M in | $/M out | Usage share |
|---|---|---|---|---|
| DeepSeek: DeepSeek V4 Flash | DeepSeek | $0.14 | $0.28 | 14.77% |
| Xiaomi: MiMo-V2.5 | Xiaomi | $0.14 | $0.28 | 8.97% |
| Tencent: Hy3 | Tencent | $0.13 | $0.53 | 7.49% |
| DeepSeek: DeepSeek V4 Pro | DeepSeek | $0.43 | $0.87 | 6.72% |
| Ling-3.0-flash (free) | Other | Free | Free | 2.77% |
| Google: Gemini 2.5 Flash Lite | $0.10 | $0.40 | 1.43% | |
| Xiaomi: MiMo-V2.5-Pro | Xiaomi | $0.43 | $0.87 | 1.16% |
| OpenAI: gpt-oss-120b | OpenAI | $0.04 | $0.17 | 0.93% |
| DeepSeek: DeepSeek V3.2 | DeepSeek | $0.27 | $0.40 | 0.80% |
| Poolside: Laguna S 2.1 | Other | $0.10 | $0.20 | 0.71% |
Price per million tokens
Cheapest output price first. Usage share shows whether anyone is actually routing to it.
Prices are OpenRouter list rates in USD per million tokens. Est. $/d = tokens × list price (30% prompt / 70% completion) and excludes volume discounts, cached tokens and free tiers.
LLM pricing FAQ
- What is the cheapest LLM API?
- Among models with measurable traffic, inclusionAI: Ling-2.6-flash (Other) has the lowest output list price at $0.03 per million tokens as of 2026-07-29. Free-tier variants aside, prices below $0.50 per million output tokens are common among open-weight models.
- Which LLM API is cheapest for real production traffic?
- DeepSeek: DeepSeek V4 Flash is the cheapest model carrying meaningful production volume: $0.28 per million output tokens at 14.77% of tracked gateway traffic.
- How is the price per million tokens calculated?
- Prices are OpenRouter list rates in USD per million tokens, split into prompt (input) and completion (output). Cached-read pricing is shown where the provider publishes it. Volume discounts and enterprise agreements are not included.
- Why compare price against usage?
- A cheap model that nobody uses tells you nothing about reliability. Pairing list price with measured gateway share shows which cheap models developers actually trust with production traffic.