Free LLM API Tokens

Where can I use DeepSeek for free via API?

14 providers in our tracker name a DeepSeek model. A provider appears here only when that name is already in our data — we do not guess the rest of a catalogue. Figures change often; confirm on the official page before you rely on them.

Last updated 13 Nov 2026. Tracker snapshot: 13 Nov 2026, 13:36 UTC.

Providers that name DeepSeek

The table lists the published free limit and OpenAI compatibility from our data. The base URL, rate limits and the clause that names this family are under each provider.

Providers in our list that name DeepSeek. Figures are copied from each provider’s own pages. “Not published” means no official number was found.
ProviderFree limitOpenAI compatible
NVIDIA Build / NIM · Official page
Base URL and extra details
Base URL
https://integrate.api.nvidia.com/v1
Named in our data as
Models include Kimi K3, DeepSeek V4.1 Flash, GLM 5.3.
Tokens/day
Not published
Limits
40 requests/min, 10,000 requests/day
Card needed
No
Status
Active
Last verified
5 Oct 2026
Free trial on hosted modelsYes
APInex · Official page
Base URL and extra details
Base URL
https://api.apinex.bond/v1
Named in our data as
Free models: free/gpt-6-luna, free/deepseek-v4.1-flash, free/gemini-3.8-flash
Tokens/day
3,000,000
Limits
5 requests/min per IP (free)
Card needed
No
Status
Active
Last verified
23 Oct 2026
3M tokens/day on free-category modelsYes
Experiential Labs · Official page
Base URL and extra details
Base URL
https://api.experientiallabs.ai/v1
Named in our data as
DeepSeek V4.1 Flash, GPT-6 Luna, Clef
Tokens/day
1,000,000
Limits
10 RPM, 250K tokens/min shared
Card needed
Yes
Status
Active
Last verified
27 Oct 2026
1M tokens/day shared on free core models (DeepSeek V4.1 Flash, GPT-6 Luna, Clef)Yes
Runcrate · Official page
Base URL and extra details
Base URL
https://api.runcrate.ai/v1
Named in our data as
DeepSeek V4.1 Flash, GLM 5.3, GLM 5.3 Flash, gpt-oss 120B, Qwen3.8 Flash
Tokens/day
Not published
Limits
Session cap plus weekly allowance (resets Monday); sizes unpublished
Card needed
No
Status
Active
Last verified
27 Oct 2026
Free usage on 5 open models (DeepSeek V4.1 Flash, GLM 5.3, GLM 5.3 Flash, gpt-oss 120B, Qwen3.8 Flash)Yes
River AI (River Inference) · Official page
Base URL and extra details
Base URL
https://api.river.ai/v1
Named in our data as
Models include DeepSeek V4.1 Flash and GLM-5.3 Flash.
Tokens/day
Not published
Limits
Not published
Card needed
Not published
Status
Active
Last verified
28 Oct 2026
$25 signup credit (launch offer)Yes
Tencent EdgeOne Makers Models · Official page
Base URL and extra details
Base URL
https://ai-gateway.edgeone.link/v1
Named in our data as
Built-in models: @makers/hy3, hy3-preview, deepseek-v4-pro, deepseek-v4-flash, deepseek-v4.1-flash, minimax-m3, minimax-m2.7, kimi-k2.6, zhuque-text
Tokens/day
Not published
Limits
300 s max per request; rate limits on abuse
Card needed
No
Status
Active
Last verified
29 Oct 2026
500K tokens/month shared across built-in models (limited-time)Yes
SenseNova Token Plan · Official page
Base URL and extra details
Base URL
https://token.sensenova.cn/v1
Named in our data as
Free public-beta Token Plan: shared credit pool across Kimi K3, GLM-5.2, DeepSeek V4.1 Flash and SenseNova models
Tokens/day
Not published
Limits
Credits per rolling 5 hours and per week (reported 120K/5h and 1.2M/week since 27 Oct); credits, not tokens
Card needed
No (mobile SMS signup)
Status
Active
Last verified
30 Oct 2026
Free public-beta Token Plan: shared credit pool across Kimi K3, GLM-5.2, DeepSeek V4.1 Flash and SenseNova modelsYes
JoinGonka Gateway · Official page
Base URL and extra details
Base URL
https://gate.joingonka.ai/v1
Named in our data as
Models: GLM-5.3-Flash, DeepSeek-V4-Flash, MiniMax-M2.7
Tokens/day
Not published
Limits
Models: GLM-5.3-Flash, DeepSeek-V4-Flash, MiniMax-M2.7; then pay per token
Card needed
No card
Status
Active
Last verified
3 Nov 2026
3M one-time starter tokens after email confirmationYes
Mervex · Official page
Base URL and extra details
Base URL
https://api.mervex.cc.cd/v1
Named in our data as
Gemma, Llama, Qwen, GLM, Kimi, DeepSeek
Tokens/day
Not published
Limits
Extra daily quota for linking Discord/Google/GitHub
Card needed
No card
Status
Active
Last verified
4 Nov 2026
Free key: 100 requests/day plus one-time +400 welcome bonus (per the team's 3 Nov launch post)Yes
Lightning AI Model APIs · Official page
Base URL and extra details
Base URL
https://lightning.ai/api/v1
Named in our data as
One key across GPT, Claude, Gemini, DeepSeek, GLM and more
Tokens/day
Not published
Limits
Free: 15 RPM, 120K tokens/min
Card needed
No
Status
Active
Last verified
5 Nov 2026
5 free credits on signup (25 more if you add a card); docs say up to 40M free tokens to start; credits expire after 12 monthsYes
OpenPaths · Official page
Base URL and extra details
Base URL
https://openpaths.io/v1
Named in our data as
Free catalogue lists many or/*:free models plus nvidia/deepseek-v4-flash and openpaths/free
Tokens/day
Not published
Limits
Best-effort rate limits per key; tokens/day unspecified
Card needed
No
Status
Active
Last verified
6 Nov 2026
Free models bill $0 per token (OpenRouter :free routes plus OpenPaths Free and others); paid models need creditsYes
Mango Inference · Official page
Base URL and extra details
Base URL
https://api.mangoboost.io/v1
Named in our data as
MangoBoost first-party OpenAI-compatible API for open-weight models such as DeepSeek V4.1 Flash, GLM-5.3, Kimi-K3
Tokens/day
Not published
Limits
Tokens/day unspecified; credits expire after stated validity period
Card needed
No
Status
Active
Last verified
6 Nov 2026
$10 free credits on signup (launch offer); 30% off models through 11 Nov 2026Yes
Zro (MoonMath) · Official page
Base URL and extra details
Base URL
https://zro.moonmath.ai/v1
Named in our data as
Free plan: daily usage on DeepSeek V4.1 Flash only
Tokens/day
Not published
Limits
Not published
Card needed
Not published
Status
Active
Last verified
12 Nov 2026
Free plan: daily usage on DeepSeek V4.1 Flash only (GLM-5.3 Flash no longer free)Yes
GonkaRouter · Official page
Base URL and extra details
Base URL
https://api.gonkarouter.io/v1
Named in our data as
GLM-5.3-Flash, DeepSeek-V4-Flash, MiniMax-M2.7
Tokens/day
Not published
Limits
Not published
Card needed
No
Status
Active
Last verified
13 Nov 2026
1M free tokens one-time for new accounts (email or Google sign-in)Yes

Example request

This example uses the published base URL for NVIDIA Build / NIM. Replace MODEL_ID with the model id from that provider’s official docs. We do not guess the id.

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["API_KEY"],
    base_url="https://integrate.api.nvidia.com/v1",
)

resp = client.chat.completions.create(
    model="MODEL_ID",  # copy the model id from the provider's docs
    messages=[{"role": "user", "content": "Hello"}],
)
print(resp.choices[0].message.content)

Common questions

Where can I use DeepSeek for free via API?

14 providers in our 13 Nov 2026, 13:36 UTC snapshot name a DeepSeek model: NVIDIA Build / NIM, APInex, Experiential Labs, Runcrate, River AI (River Inference), Tencent EdgeOne Makers Models, SenseNova Token Plan, JoinGonka Gateway, Mervex, Lightning AI Model APIs, OpenPaths, Mango Inference, Zro (MoonMath), GonkaRouter. We only list a provider here when that name already appears in our data. Catalogues change, so confirm on the official page before you rely on an entry.

Which of these DeepSeek providers are OpenAI compatible?

14 of the 14 active entries are marked OpenAI compatible in our data: NVIDIA Build / NIM, APInex, Experiential Labs, Runcrate, River AI (River Inference), Tencent EdgeOne Makers Models, SenseNova Token Plan, JoinGonka Gateway, Mervex, Lightning AI Model APIs, OpenPaths, Mango Inference, Zro (MoonMath), GonkaRouter. The others are marked no, partial, or Not published — see the table on this page.

Do I need a credit card?

11 active entries in this list say no card is required: NVIDIA Build / NIM, APInex, Runcrate, Tencent EdgeOne Makers Models, SenseNova Token Plan, JoinGonka Gateway, Mervex, Lightning AI Model APIs, OpenPaths, Mango Inference, GonkaRouter. Others may ask for a card, a top-up, or do not publish this. Confirm on the sign-up page.

Why do some fields say Not published?

The provider has not published that figure on their public pages, or we could not find it. We never invent quotas, prices or endpoints.

Is this every site that hosts DeepSeek?

No. This page only includes providers that already name DeepSeek in our data file. A gateway may offer the model without that name appearing in our notes, allowance or limits, and we do not guess those catalogues.