Relace on Concentrate: DeepSeek, GLM and Kimi With No Fees
Relace is live on Concentrate: DeepSeek V4.1, GLM-5.3 and Kimi K3 from $0.07/M input tokens, with no credit fees, no per-request fees and no token markup.
Concentrate.ai
On this page
Relace is now live on Concentrate. You can call Relace's DeepSeek V4.1, DeepSeek V4 Pro, GLM-5.3 and Kimi K3 routes through one Concentrate API key, at Relace's listed token prices, with no fee on credit purchases, no per-request fees and no markup on tokens. Prices start at $0.07 per million input tokens.
What is Relace?
Relace is an inference provider for open-weight models, and one of the highest-volume providers on any AI gateway. According to Relace, it served 30 trillion tokens in the 30 days before its latest update and ranked first among OpenRouter providers by daily tokens on October 5, 2026 (2.3 trillion tokens that day). Relace says it is usually the cheapest option for the models it serves.
On Concentrate, every Relace route is marked Zero Data Retention, and all ten models have a 1M token context window.
Which Relace models can I use on Concentrate?
Ten models from DeepSeek, Z.ai (GLM) and Moonshot (Kimi) are live on Relace today:
Model
Concentrate model id
Input $/M
Output $/M
LLM Gateway
Instant access to every model
Change models, providers, add fallbacks, and more by running all AI usage through one platform.
Prices as of November 2026. See the Relace provider page for live prices, cache pricing and new models as Relace adds them.
How much does Relace cost through Concentrate?
You pay Relace's token prices and nothing on top. Concentrate charges no fee when you buy credits, no per-request fee and no markup on tokens. Cached input is cheaper still: on DeepSeek V4.1 Flash 1017, Relace lists cache reads at $0.01 per million tokens.
That matters most at Relace's price points. When input costs $0.07 per million tokens, a percentage fee on every credit top-up is a real share of your bill.
How do I call Relace models through Concentrate?
Use the provider-prefixed model id to pin a request to Relace:
curl https://api.concentrate.ai/v1/responses \
-H "Content-Type: application/json" \
-H "Authorization: Bearer YOUR_API_KEY" \
-d '{
"model": "relace/deepseek-v4-1-flash-1017",
"input": "Summarize this support ticket in two sentences."
}'
Or send the model name without a prefix and let Concentrate route it. Setting "routing": {"provider": {"sort": "cost"}} sends the request to the cheapest provider serving that model, and if a provider goes down, the request reroutes to another one. Pick a model. Pick your policies. Fallbacks are built in. Full reference in the Concentrate docs.
How does Concentrate compare with OpenRouter and Vercel AI Gateway for Relace?
Relace is available on all three gateways, and all three pass through provider token prices. The difference is in the fees around them.
One API key for everything: Relace sits next to about 200 models from about 30 providers, including Claude, GPT and Gemini, so you can mix cheap open-weight calls with frontier models without another account.
Spend limits: issue virtual keys per team, app or customer, each with its own budget, and track spend in real time.
Automatic rerouting: many of these models are served by several providers on Concentrate (DeepSeek V4.1 Flash 1017 has eight), so traffic keeps flowing if one has an outage.
Data redaction: filter PII out of prompts before a request leaves Concentrate.
Can I use Relace with patient data (PHI)?
Not today. Relace routes are not BAA-eligible on Concentrate. For PHI, use the BAA-eligible routes listed on the HIPAA LLM gateway page, including DeepSeek and GLM on other providers; see DeepSeek V4.1 Flash with PHI and GLM-5.3 Flash with PHI. With a BAA, Concentrate is $499/month, with no per-request fees and no markup on tokens.
Add credits. There is no fee on top of what you load.
Call any model on the Relace provider page with its relace/ model id, or let cost routing pick Relace when it is the cheapest.
FAQ: Relace on Concentrate
Is Relace available on Concentrate?
Yes. As of November 5, 2026, ten Relace models are live on Concentrate, covering DeepSeek V4.1 Flash, DeepSeek V4 Pro and Flash, GLM-5.3, GLM-5.3-Flash, GLM-5.2 and Kimi K3.
Does Concentrate charge a fee on Relace tokens?
No. You pay Relace's listed token price. Concentrate adds no markup, no per-request fee and no fee on credit purchases.
What is the cheapest model on Relace?
DeepSeek V4 Flash (0731 and 0423) at $0.07 per million input tokens and $0.14 per million output tokens on Concentrate as of November 2026.
Does Relace keep my prompts?
Every Relace route on Concentrate is marked Zero Data Retention. Check the badge on each model page for the current status.
Can I use Relace models with PHI?
No. Relace routes are not BAA-eligible on Concentrate today. Use the BAA-eligible routes on the HIPAA LLM gateway page instead.
How is this different from using Relace on OpenRouter?
Both pass through Relace's token prices. OpenRouter charges a 5.5% fee when you buy credits on its Standard plan; Concentrate charges no fee on credit purchases and no per-request fees.
Can I call Relace directly instead?
Yes, Relace sells its own API. Going through Concentrate adds one key for Relace plus about 30 other providers, automatic rerouting to another provider if Relace has an outage, and spend limits per key.
Blog