Telluvian vs Not Diamond: AI model routers compared
Telluvian and Not Diamond both predict which model will answer a prompt best. The main difference is what happens next: Not Diamond returns a recommendation and you call the provider yourself, while Telluvian picks the model and returns its answer in the same OpenAI-compatible request, with optional per-token hallucination scores. Telluvian also offers the recommendation on its own, through /v1/modelSelect.
Last reviewed
At a glance
| Telluvian | Not Diamond | |
|---|---|---|
| Model selection method | Prompt-aware. Model Select predicts each model’s performance on the specific prompt and picks the cheapest one that clears your xPerf quality bar. | Prompt-aware. A pre-trained router, or a custom router trained on your evaluation data, predicts the best model per query in quality, cost or latency mode. It returns the recommendation; you call the model yourself. |
| Models and providers | 190+ models from OpenAI, Anthropic, Google, Qwen, DeepSeek, xAI, Z-AI, Moonshot and Meta, behind one OpenAI-compatible API. | 90+ models from 13 providers including OpenAI, Anthropic, Google, Mistral, xAI and DeepSeek. Custom and fine-tuned models through a custom router. |
| Pricing model | Prepaid, pay as you go. Provider list price for tokens, plus $0.05 per 1M input tokens for Model Select and $1.00 per 1M completion tokens for hallucination scores (optional). | $0.05 per million tokens routed on pay as you go; volume discounts on Enterprise. You pay model providers directly, since you make the model call. |
| Hallucination detection | Built in. Per-token hallucination scores from probes reading an open-weight proxy model, including for closed models. Off by default. | None documented. |
Key differences
Integration
Telluvian returns the model's answer; Not Diamond returns a recommended model that you then call with your own SDK.
Billing
with Telluvian, routing and model tokens come from one prepaid balance; with Not Diamond you pay Not Diamond and each provider separately.
Quality bar
Telluvian takes xPerf, a model name or a number on a fixed scale; Not Diamond optimises in quality, cost or latency mode.
Custom routers
Not Diamond can train a router on your evaluation data; Telluvian's Model Select works from the first request without it.
Hallucination detection
Telluvian can score every generated token; Not Diamond does not document an equivalent.
Frequently asked questions
Does Not Diamond call the model for me?
No. Not Diamond returns the recommended model and its quickstart then has you call that model with your own SDK. Telluvian's Gallery does both in one request. If you prefer the recommendation pattern, Telluvian's /v1/modelSelect endpoint returns only the choice.
How does pricing compare?
The routing fee is the same headline rate: Not Diamond charges $0.05 per million tokens routed, and Telluvian $0.05 per 1M input tokens when Model Select picks the model. With Telluvian the routing fee and the model tokens come out of one prepaid balance at provider list price; with Not Diamond you pay Not Diamond and each provider separately.
Do I need my own data to start routing?
Not with Telluvian. Model Select works from the first request: set xPerf to the standard you need and, if you like, limit the candidates with modelZoo. Not Diamond's custom routers need your evaluation data, though its pre-trained router does not.
Does Not Diamond detect hallucinations?
We could not find hallucination detection in Not Diamond's docs. Telluvian returns a hallucination score for every generated token when you set include_scores to true.
Does Telluvian's quality bar change when new models launch?
No. xPerf is measured against a fixed reference at 1.0, so 0.9 today means 0.9 next year, and new models are placed on the same scale. You can also name a model as the bar. Not Diamond instead asks you to choose a quality, cost or latency mode.
Try it on your own prompts
Point an OpenAI SDK at Telluvian, send telluvian/gallery-1 as the model, and see which model answers each request. Questions about your use case go straight to the team.
Sources
Competitor details come from their own public docs and pricing pages, checked on . Where a page did not say, neither do we. Spotted something out of date? Tell us at hello@telluvian.ai.