Telnyx Inference API

What it does

Telnyx Inference API takes chat completion requests over HTTP, including messages, tool definitions, and structured output settings, and returns model responses such as text, streamed tokens, tool calls, or embeddings. Developers call it by changing the base URL and model name in an existing OpenAI SDK client, and choose among a hosted library of open-weight models running on Telnyx-operated GPUs.

People use it when they want model serving behind a standard API rather than a chat interface, such as adding summarization, classification, or agent reasoning to an application. Distinctive points on the product page are in-region inference across the Americas, Europe, MENA, and APAC, a zero data retention statement, and service tiers (Default, Priority, Flex) that change serving capacity and rate rather than the model.

It suits developers and engineering teams building production applications who already use OpenAI-compatible tooling. Access requires a Telnyx account and API key. Limitations include that Priority and Flex are available only on select models, Flex is US-only on select models, and published benchmark results are specific to the tested model and workload and are not an SLA.

What it can do

• OpenAI-compatible chat completions endpoint usable from existing OpenAI SDKs by changing the base URL
• Streaming responses and tool calling on supported models
• Structured output via JSON mode and regex constraints
• Embedding models for search and RAG, such as Qwen3-Embedding-8B
• Service tiers Default, Priority, and Flex with different serving capacity and rates
• Autoscaling of concurrent requests on Telnyx-owned GPU infrastructure

Works with

• OpenAI SDK compatibility via base URL change

Before you choose

• Priority and Flex tiers are available only on select models; Flex is US-only on select models.
• Published performance results are specific to the tested model and workload and are not an SLA.
• Per-token rates vary by model and service tier; confirm the current rate table before estimating costs.

Listed under

• Coding
• Sandboxes and infrastructure

Get started

Sign up, then create an API key and make a test request such as GET /v2/balance.

1. Sign up for a Telnyx account and create an API key.
2. Export the key as TELNYX_API_KEY and validate it with a read-only GET /v2/balance request.
3. Set the Telnyx base URL, API key, and a supported model in your existing OpenAI-compatible client and send a chat completion request.

Pay-as-you-go per-token billing with no minimums on standard rates, a card on file, and automatic volume tier discounts. Sign-up free allowance amount is not stated. The enterprise tier listed on the page has a $5,000/month minimum spend, which is a separate plan. Volume rates require contacting sales.

Source: telnyx.com/pricing/inference-api, checked 29 Oct 2026. Prices can change; check before you buy.

Sources for this overview

Product pages, documentation, and pricing checked 29 Oct 2026. Each source supports part of the overview; availability and prices can change.

• Run the best open-source models. One Inference API. ↗telnyx.com
• Infrastructure for realtime agents | Telnyx ↗telnyx.com
• Telnyx Developer Docs - Telnyx ↗developers.telnyx.com
• Inference Pricing | Telnyx ↗telnyx.com

Company website

telnyx.com: product, pricing and policy pages

Social accounts

Accounts the company links, and its posts on X

Human review

Checked by a person

knowledge_base

manual_review

Last checked 29 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.