We've expanded the gateway to the newest models from both major providers. Nothing changes in your integration — same key, same endpoint. You just pass a new model name.
Key takeaways
- Full GPT-6 (Astra) + GPT-5.x line and the Claude 5 family, all on one key.
- Embeddings route through the gateway too — one bill for chat and vectors.
- Every model is provider list price plus a flat 25%, with a monthly spend cap.
The OpenAI line
GPT-6 Astra sits at the top as the flagship, with the GPT-5.6 family (Sol, Terra, Luna) and the rest of the 5.x line beneath it for cost- and latency-sensitive work.
| Tier | Models | Best for |
|---|---|---|
| Flagship | GPT-6 Astra | Hard reasoning, long context, agents |
| High-end | GPT-5.6 Sol / Terra | Strong general work at lower cost |
| Fast | GPT-5.6 Luna, 5.x mini / nano | High-volume, latency-sensitive paths |
The Claude 5 family
Anthropic runs as a real passthrough — including tool calling — so Claude behaves natively, not as a lowest-common-denominator shim.
| Model | Best for |
|---|---|
| Claude Opus 5 | Deepest reasoning, long agentic runs |
| Claude Sonnet 5 | The balanced everyday workhorse |
| Claude Fable 5.1 / Haiku 4.5 | Fast, cheap, high-throughput |
Embeddings, too
text-embedding-3-small and -3-large now route through the gateway, so your vector pipeline shares the same key and bill as chat.
Nothing changes in your code
If you already call the gateway, adding a model is a one-line change — swap the string:
curl https://withconflux.com/api/v1/chat/completions \
-H "Authorization: Bearer $CFLX_KEY" \
-d '{"model":"gpt-6-astra","messages":[{"role":"user","content":"Hello"}]}'The model slug you pass is the real provider model ID — so anything the provider ships, you can call the moment we add it to the catalog.
One integration, every model. Pick the right one per workload — the full list and live pricing is here.