Free LLM API Credits: Give Your Companion a Brain for $0
The same AniMate companion can feel like two completely different characters depending on the model behind her. The good news: in 2026 nearly every major model provider is handing out free credits, permanently free models, or referral bonuses — and getting a key usually takes about three minutes. Here's how to pick one and get chatting without spending anything.
Three things to weigh before picking a chat model
1. Latency and your network
A desktop companion should answer while the moment is still alive. Latency mostly comes down to where the servers are. If you're in China, providers with domestic nodes — Zhipu, Xiaomi MiMo, SiliconFlow, Qwen — give you the fastest first token. Models like Gemini, Claude, or OpenAI depend on how stable your network route to them is; a flaky route means stutters and dropped replies. The same logic applies in reverse for users in Japan, Korea, Europe, or the US: pick providers close to you.
2. Quality — how well she stays in character
Your companion's personality lives in the conversation: keeping a long persona consistent, catching your jokes, holding a stable tone. Flagship models (GLM-5.3, MiMo V2.5, Gemini, Claude) are noticeably better at long-context persona memory, emotional tone, and multilingual chat. Free smaller models handle casual chat fine, but complex personas will drift sooner. A good strategy: start free to get things running, spend referral credits on a flagship to feel the difference, then decide what's worth paying for.
3. Cost — where the genuinely free models are
Three ways to keep the bill at zero, from most to least durable:
- Permanently free models: Zhipu's GLM-4.7-Flash, SiliconFlow's zero-cost model tier, and OpenRouter's
:freemodels — these aren't promotions, they're standing offers. - Sign-up and referral credits: Xiaomi MiMo (¥10 for both sides), Zhipu's invite bundle (20M tokens), SiliconFlow's ¥16 voucher after identity verification, Qwen AI's new-user quota.
- Dirt-cheap fallback: DeepSeek has no free tier, but its pay-as-you-go pricing is extremely low — a small top-up lasts months of desktop-companion chat.
Quick picks by region
For users in China
- Start at $0: SenseNova's Token Plan (fully free, three models including DeepSeek V4 Flash) and Zhipu GLM (
glm-4.7-flash, free, phone-number sign-up). - Stack the bonuses: Xiaomi MiMo (¥10 for both sides via invite) + SiliconFlow (¥16 voucher) + Qwen AI (new-user quota) — register all three and the combined free credits last a month or two of heavy use.
- Long-term fallback: DeepSeek — top up via WeChat/Alipay, rock-bottom pricing.
For users in Japan and Korea
- Start here: Google Gemini's free tier (key in one click via AI Studio, no card, generous daily quota, reliable access and strong multilingual quality).
- Model variety: OpenRouter's
:freemodels — try DeepSeek, Qwen, Llama, and GLM free versions side by side. - Flagship taste: Claude (~$5 trial credit, phone verification required) or Mistral's Experiment tier (all models free at ~1 request/second).
For users in Europe and the US
- Start at $0: Gemini's free tier, or Groq / Cerebras — both OpenAI-compatible with no card required (Cerebras gives away 1M tokens a day; Groq's free tier allows thousands of requests daily). Connect them through AniMate's OpenAI Compatible preset.
- Widest free catalog: OpenRouter
:freemodels — 50 requests/day on a free account, 1,000/day after a one-time $10 top-up (credit never expires). - Flagship taste: Claude's ~$5 trial credit (phone verification), or xAI Grok's $25 sign-up credit when available.
Provider-by-provider reviews
Everything below is bundled as a one-click preset in AniMate's Add LLM Service dialog — pick the provider and the base URL and recommended model fill themselves in. SenseNova isn't a built-in preset yet, but it's fully OpenAI-compatible and free, so it earns a spot here too.
SenseNova: completely free, the best pure bargain
Free tier: The Token Plan is free with no top-up and no card — three models (DeepSeek V4 Flash, SenseNova 6.7 Flash-Lite, SenseNova U1 Fast) with separate quotas of 1,500 calls every 5 hours each. Desktop-companion chat won't come close to that limit.
Sign-up experience: Phone number, then create a key at platform.sensenova.cn/console/keys (up to 20 keys per account).
Heads-up: The free plan is a limited-time public beta; paid tiers are coming, so keep a second provider configured as backup. One account covers both DeepSeek V4 Flash and SenseNova's own models.
- Connecting to AniMate: Add LLM Service → provider "OpenAI Compatible" → base URL
https://api.sensenova.cn/v1→ paste your key → model ID for DeepSeek V4 Flash or a SenseNova model (check the console's model list for exact IDs).
Zhipu GLM: the free workhorse — start here
Free tier: GLM-4.7-Flash is completely free, no top-up needed (the platform lists seven free models); the flagship GLM-5.3 is pay-as-you-go.
Sign-up experience: Phone number, create a key, done — no card, domestic latency, the smoothest possible start.
Heads-up: Free models are rate-limited per minute. Fine for a single companion, noticeable if you run several characters at once; the free model list lives under the "free" filter in the model catalog.
Xiaomi MiMo: the most useful credits — ¥10 for both of you
Free tier: ¥10 in trial credits for new accounts (valid 40 days), plus the "MiMo 100T" application program with up to 100 trillion tokens for approved applicants. MiMo V2.5 runs a 128K context window and holds a character persona well.
Sign-up experience: Log in with a Xiaomi account, no card. The same platform also hosts MiMo's TTS models, which AniMate supports — one account covers chat and voice.
Heads-up: Trial credits expire after 40 days; referral rewards only apply when signing up through an invite link.
SiliconFlow: lots of free models + a voucher
Free tier: A tier of zero-cost models (small Qwen variants and more); a ¥16 voucher after identity verification under "verification perks."
Sign-up experience: Phone number registration; filter the model catalog by "free." AniMate's preset defaults to Qwen/Qwen3-8B.
Heads-up: The free model list and bonus amounts change often — confirm in the billing center. The voucher only appears after completing identity verification.
Qwen AI (Alibaba): the easiest Alibaba entry point
Free tier: New users get free quota for the Qwen model family (exact amounts shown at sign-up); users with a verified Alibaba Cloud account can also claim 1M tokens per model (90 days) on the Model Studio console.
Sign-up experience: platform.qianwenai.com is the simplest on-ramp in Alibaba's ecosystem — a few clicks and you have a key.
Heads-up: Free quota is time-limited; turn on the spend cap. qwen-flash is AniMate's preset default and works out of the box.
DeepSeek: no free tier, but cheap enough to be your fallback
Free tier: None — pure pay-as-you-go. v4-flash costs about $0.66 per million output tokens off-peak; $5 of credit goes a long way.
Sign-up experience: Phone number, top up with WeChat or Alipay, no card.
Heads-up: Watch for peak/off-peak pricing differences; earlier promotional discounts have ended.
MiniMax: versatile domestic option
Free tier: New-user credits (amount shown in your console); the M2.7 Highspeed model targets low latency, which suits real-time desktop chat.
Sign-up experience: Phone registration; the platform also hosts MiniMax TTS, which AniMate supports.
Heads-up: Some free endpoints are limited to early users; confirm current bonuses in the console.
Google Gemini: the most generous standing free tier
Free tier: Free in AI Studio with no card — Flash-Lite at roughly 1,500 requests/day, Flash at 250–500/day.
Sign-up experience: Google account, one click to create a key. The easiest free option outside China.
Heads-up: Quotas are per project and shift often; not directly reachable from mainland China without your own network arrangement.
OpenRouter: one key, every free model
Free tier: Every :free model costs nothing; 50 requests/day on a free account, 1,000/day after a one-time $10 top-up that never expires.
Sign-up experience: Email registration, create a key, go.
Heads-up: Free models cap output length per request and queue behind paid traffic at peak times.
Mistral: every model, free tier
Free tier: The Experiment plan unlocks all models at about 1 request/second, no credit card.
Sign-up experience: Email registration.
Heads-up: Positioned for testing and prototyping; the rate limit can change without notice.
Anthropic Claude: $5 to taste a flagship
Free tier: A one-time ~$5 trial credit across all Claude models.
Sign-up experience: No card, but phone verification is mandatory — some numbers fail, and without it the credit never lands.
Heads-up: $5 is roughly a day of light flagship use; treat it as a quality benchmark run.
xAI Grok: sign-up credit, shifting terms
Free tier: Widely reported $25 for new accounts (valid ~30 days), though xAI doesn't officially advertise it — check the sign-up page.
Sign-up experience: Email registration; a data-sharing program offers $150/month in credits but requires a card and irrevocable consent — skip it if that bothers you.
Heads-up: Bonus terms change frequently; don't build your setup around them.
OpenAI: no free tier
Free tier: The old $5 grant has largely been retired; pay-as-you-go with a card is the only route. Not the first stop for a zero-cost setup.
Plug the key back into AniMate
- Open Settings → AI Services in AniMate and click "Add LLM" under LLM services.
- Pick the provider from the dropdown — base URL and recommended model fill in automatically.
- Paste your API key, check the model menu, save, and hit "Test connection."
- When you see a latency number, you're connected — set it as default and start chatting.
FAQ
What happens when the free credits run out?
Register somewhere else and claim a new batch, or keep several configured and switch the default in AniMate whenever one dries up. DeepSeek's low pay-as-you-go rates make it a solid long-term fallback.
Can I configure multiple providers at once?
Yes — AniMate supports multiple LLM connections with independent providers, keys, and model menus, plus one-click default switching.
Are the referral bonuses real?
They come from each platform's current invite program (Xiaomi's ¥10 for both sides, Zhipu's 20M-token bundle, SiliconFlow's ¥16 voucher). Programs change, so treat the sign-up page as the source of truth.
Is my API key safe?
Keys live in a local config folder on your machine. Chat requests go directly from your PC to the provider — they never touch AniMate's servers.
Give her a brain in three minutes
Claim any provider's free credits, paste the key into AniMate, and start talking. Free speech-to-text and text-to-speech credits are covered in the next two guides.