Access GPT, Claude, Grok, Qwen and more through one OpenAI-compatible endpoint — without separate subscriptions.
Three ready packages, or set an exact amount. Pay with PayPal Friends & Family or crypto (USDT, Bitcoin, Litecoin) — keys are issued on Discord, usually in 5–15 min. Totals below already include the 3% method fee.
Enough to connect Cursor, test models and ship a first project.
Buy StarterDaily coding volume. Volume pricing — larger volumes cost less per 1M, same curve as the slider.
Buy BuilderLarge prepaid balance for sustained work across several models.
Buy HeavySame models, published output-token list prices against the NullRoute estimate. This is the 97% figure — not a slogan on its own.
| Model | Direct API | NullRoute | Saving |
|---|---|---|---|
| GPT 6 Astra | $25.00 / 1M output | ≈ $0.75 | 97% |
| Claude Fable 5.1 | $25.00 / 1M output | ≈ $0.80 | 97% |
| Grok 4.6 | $6.00 / 1M output | ≈ $0.10 | 98% |
Direct prices are published output-token references in USD. NullRoute figures: cost = 1M ÷ usage rate × ~€0.08 / 1M base, shown in USD at face value. Provider pricing and availability can change.
One base URL and one key. If a client accepts a custom OpenAI endpoint, it works.
Settings → Models → Override OpenAI Base URL, paste the URL below and your nr- key.
Set base_url to the NullRoute endpoint and api_key to your key. Same key as the dashboard.
Any OpenAI-compatible extension → custom base URL + key. Takes under a minute.
OpenAI client, LangChain, or raw HTTP. Authorization: Bearer nr-...
base_url: https://api.getnullroute.com/v1 api_key: nr-... (your key from Discord) model: gpt-6-astra
Choose volume, receive a key, connect it. No separate lab accounts.
Starter, Builder, Heavy, or a custom amount on the slider.
Pay on Discord. You get a private nr- key and dashboard access.
Paste the base URL and key into Cursor, Codex, VS Code or your SDK.
base_url: https://api.getnullroute.com/v1 api_key: nr-...
You buy a pool of NullRoute tokens. Each model uses that pool at a different rate. Lower rates make the same balance last longer.
Usable tokens = volume you bought ÷ usage rate
Rate ×0.1 — DeepSeek V4 Flash. Every 1 token bought gives you 10 tokens to run.
Rate ×7.5 — GPT 6 Astra. Uses more of the pool, still about 97% under official output-token list.
Usage rate shows how quickly a model spends your balance. Lower is cheaper to run. Click a row to copy its model ID. AA index is an Artificial Analysis reference, not a universal quality score.
| Model | ID | Usage rate | AA index |
|---|---|---|---|
| GPT 6 Astra | gpt-6-astra | ×7.5 | 52.8 |
| GPT 5.6 Sol | gpt-5.6-sol | ×3 | 47.0 |
| GPT 5.6 Terra | gpt-5.6-terra | ×1.5 | 42.2 |
| GPT 5.6 Luna | gpt-5.6-luna | ×0.83 | 37.9 |
| GPT 5.5 | gpt-5.5 | ×3 | 40.1 |
| Claude Fable 5.1 | claude-fable-5-1 | ×8 | 53.4 |
| Claude Fable 5 | claude-fable-5 | ×8 | 49.7 |
| Claude Opus 5 | claude-opus-5 | ×4 | 51.3 |
| Claude Opus 4.8 | claude-opus-4-8 | ×4 | 42.2 |
| Claude Opus 4.7 | claude-opus-4-7 | ×4 | 40.1 |
| Claude Opus 4.6 | claude-opus-4-6 | ×4 | 37.9 |
| Claude Sonnet 5 | claude-sonnet-5 | ×2 | 37.9 |
| Claude Sonnet 4.6 | claude-sonnet-4-6 | ×2 | 30.4 |
| Claude Haiku 4.5 | claude-haiku-4-5 | ×0.9 | 18.2 |
| GLM 5.3 | glm-5.3 | ×1.5 | 44.9 |
| GLM 5.3 Flash | glm-5.3-flash | ×0.3 | 42.2 |
| GLM 5 Turbo | glm-5-turbo | ×1 | 27.8 |
| GLM 5.2 | glm-5.2 | ×1.5 | 34.2 |
| Grok 4.6 | grok-4.6 | ×0.5 | 43.8 |
| Grok 4.5 | grok-4.5 | ×0.5 | 39.0 |
| Composer 2.5 Fast | composer-2.5-fast | ×0.3 | 35.8 |
| Qwen 3.8 Flash | qwen3.8-flash | ×0.7 | 40.1 |
| Qwen 3.8 Max | qwen3.8-max | ×2.5 | 53.4 |
| Kimi K3 | kimi-k3 | ×2.5 | 43.8 |
| Kimi K2.7 Code | kimi-k2.7-code | ×0.9 | 26.2 |
| Muse Spark 1.3 | muse-spark-1.3 | ×0.07 | 48.0 |
| Hy 4 Preview | hy4-preview | ×1 | 34.2 |
| DeepSeek V4 Pro | deepseek-v4-pro | ×0.5 | 35.8 |
| DeepSeek V4.1 Flash | deepseek-v4.1-flash | ×0.3 | 40.1 |
| DeepSeek V4 Flash | deepseek-v4-flash | ×0.1 | 35.2 |
| MiniMax M3 | minimax-m3 | ×0.3 | 29.9 |
| MiMo V2.5 Pro | mimo-v2.5-pro | ×0.3 | 26.2 |
| MiMo V2.5 | mimo-v2.5 | ×0.05 | 21.9 |
| Gemini 3.1 Pro | gemini-3.1-pro | ×2 | 29.9 |
| Gemini 3.6 Flash | gemini-3.6-flash | ×2 | 35.8 |
| Gemini 3.7 Flash | gemini-3.7-flash | ×2 | 37.9 |
| Gemini 3.8 Flash | gemini-3.8-flash | ×2 | 41.1 |
No models match that filter.
Usage rate: lower rates make your balance last longer. Benchmark scores are from Artificial Analysis and are not universal quality guarantees. Index values are shown on the AA scale (Claude Fable 5.1 = 53.4), not as a percentage of “overall quality.”
Trust first. Adjectives later.
How to buy, how billing works, and what happens when a call fails.
Pick Starter, Builder, Heavy, or a custom volume, then open checkout. There is no card form on this site. You join Discord, send the package, pay, and receive a private NullRoute key.
Yes. Use the OpenAI-compatible endpoint. Set base_url to https://api.getnullroute.com/v1 and the API key from your account. Setup typically takes under a minute.
Your key can use the 37 models listed under Models. Availability can change. You select a model by its ID in the request.
You select the model by name. NullRoute is not affiliated with OpenAI, Anthropic, or xAI. Availability can change if a route is withdrawn.
No. 100M is NullRoute token volume. Usable tokens = volume ÷ usage rate — see the math section above. The full comparison table is under Compare.
Usage rate is how fast a model spends your NullRoute balance. Lower rates make the same top-up last longer. It is not a quality score.
No. Unused volume is not refundable. A refund applies only if the account cannot be used or paid tokens never appear after payment.
Technical logs are kept 24h for metering and abuse prevention, then deleted — never used for training. Requests are forwarded to the model provider, which applies its own rules. Do not include secrets in prompts.
Failed requests are not billed. Rare anti-abuse rule, defined in the Offer: only sustained output-heavy traffic (over 10× more output than input across many requests) can be billed at ×3. Ordinary coding, chat and agent use is never affected.
Starter is €8.24 total. PayPal F&F or crypto on Discord, key in ~5–15 min. Then paste it into Cursor, Codex, or your SDK.