OpenRouter OpenAI-compatible
Router across 200+ models — one key, free tier included.
OpenRouter is a routing layer that exposes ~200 models from dozens of upstream providers behind a single OpenAI-compatible endpoint. About 25 of them are flagged `:free` — usable at no per-token cost up to a daily request limit. With $10+ in account credit the daily free-tier limit jumps from ~50 to ~1000 requests/day, which is the threshold that makes them practical for benchmarking.
Strengths
- One key, ~200 models — including 25 free `:free` variants
- $10+ credit unlocks ~1000 free-model requests/day
- Useful for cross-provider drift checks (same model, different routes)
When to use it
- Exploring models you haven't committed to a provider for
- Comparing how the same open-weight model behaves across providers
- Light/occasional usage where the free tier suffices
Subscription plans
| Plan | Price | Quota | Available |
|---|---|---|---|
| Free tier | $0/mo | ~50 `:free` requests/day (rises to ~1000/day with $10+ credit) | yes |
| Pay-as-you-go | $0/mo | Per-token pricing on non-free models · deposit any amount | yes |
Notes: OpenRouter has no subscription tiers — `:free` model quotas are tied to your credit balance, not a recurring plan. Several `:free` models share rate limits with their upstream provider; a few have been returning HTTP 429s persistently regardless of OpenRouter quotas.
Models tested on OpenRouter
Speed numbers below are specific to OpenRouter's routing and hardware. The same model may appear on other providers' pages with different throughput.
| Model | Best tok/s | Avg tok/s | Runs | Success | Longest output (chars) |
|---|---|---|---|---|---|
| nvidia/nemotron-3-nano-30b-a3b:free | 4177.4 | 4177.4 | 1 | 100% | 5,886 |
| poolside/laguna-xs.2:free | 3746.4 | 3746.4 | 1 | 100% | 2,230 |
| liquid/lfm-2.5-1.2b-thinking:free | 2416.9 | 2416.9 | 1 | 100% | 2,098 |
| openai/gpt-oss-20b:free | 1737.2 | 1737.2 | 1 | 100% | 4,823 |
| liquid/lfm-2.5-1.2b-instruct:free | 1115.9 | 1115.9 | 1 | 100% | 1,973 |
| nvidia/nemotron-nano-9b-v2:free | 1018.1 | 1018.1 | 1 | 100% | 767 |
| nvidia/nemotron-nano-12b-v2-vl:free | 946.8 | 946.8 | 1 | 100% | 2,687 |
| google/gemma-4-31b-it:free | 945.2 | 945.2 | 2 | 50% | 3,484 |
| minimax/minimax-m2.5:free | 857.1 | 857.1 | 2 | 50% | 1,887 |
| openai/gpt-oss-120b:free | 332.9 | 332.9 | 1 | 100% | 4,263 |
| google/gemma-4-26b-a4b-it:free | 284.5 | 284.5 | 3 | 33% | 3,393 |
| z-ai/glm-4.5-air:free | 187.8 | 187.8 | 1 | 100% | 3,206 |
| nvidia/nemotron-3-nano-omni-30b-a3b-reasoning:free | 29.1 | 29.1 | 1 | 100% | 0 |
| poolside/laguna-m.1:free | 3.5 | 3.5 | 1 | 100% | 0 |