Best overall free frontier
NVIDIA NIM — poolside/laguna-xs-2.1, 262K ctx · OpenRouter — nvidia/nemotron-3-ultra-550b-a55b:free, 1M ctx
The free tier index.
Verified, not scraped.
30 providers, ~429 free models. Status comes from live probes and official docs as of 2026-09-11 — not aggregator claims. When a free tier dies, it gets marked here.
0providers tracked
0free models
29/30no credit card
0at-risk (0 online)
Status legend
All providers, ranked by current health.
🟢 Confirmed at last verification🟡 Tier exists, limits/models shifted⚠️ Reported free, 0 models online⚪ Reported free, not yet verified
| Provider | Status | Free models | Free tier | Base URL |
|---|---|---|---|---|
| Groq groq/compound (30/250) · openai/gpt-oss-120b (30/1K) · qwen/qwen3.6-27b · llama-prompt-guard-2 (30/14.4K) | 🟢 FREE | 13 | Permanent free; ~30 RPM / 1K RPD typical, prompt-guard models up to 14.4K RPD probe: live · 14 models · 2026-09-11 updated 2026-08-25 | https://api.groq.com/openai/v1 |
| Google Gemini (AI Studio) gemini-3.6-flash (15/1500) · gemini-3.5-flash (15/1500) · gemini-3.5-flash-lite (30/1500) | 🟢 FREE | 17 | Permanent free; 15–30 RPM / 1,500 RPD on Flash models updated 2026-08-24 | https://generativelanguage.googleapis.com/v1beta |
| NVIDIA NIM poolside/laguna-xs-2.1 (262K) · z-ai/glm-5.1 (202K) · glm-5.2 removed (410) | 🟢 FREE | 126 | 126 free models, permanent; 40 RPM, NO daily cap — most forgiving app tier probe: live · 80 models · 2026-09-11 updated 2026-08-25 | https://integrate.api.nvidia.com/v1 |
| OpenRouter nvidia/nemotron-3-ultra-550b-a55b:free (1M) · z-ai/glm-5.2:free (256K) · poolside/laguna-*-2.1:free · openrouter/free router | 🟢 FREE | 15 :free | ~28 :free models; free tier 50 req/day, $10 topup → 1K RPD probe: live · 444 models · 2026-09-11 updated 2026-09-11 | https://openrouter.ai/api/v1 |
| Cerebras gpt-oss-120b · gemma-4-31b (live catalog) | 🟢 FREE | 2 (gated) | 2 models in catalog (gpt-oss-120b, gemma-4-31b); inference gated by billing setup probe: live · 3 models · 2026-09-11 updated 2026-08-25 | https://api.cerebras.ai/v1 |
| Mistral AI labs-leanstral-2603 · mistral-moderation-2603 (free-tagged) | 🟢 FREE | 12 | ~1B tokens/month experiment plan; free tier does NOT log prompts updated 2026-08-24 | https://api.mistral.ai/v1 |
| Ollama Cloud minimax-m3 · gpt-oss:20b · nemotron-3-ultra | 🟢 FREE | 13 | 13 free models; session/weekly limits updated 2026-08-24 | https://api.ollama.com |
| GitHub Models Phi-4 · Mistral-large-2411 · AI21-Jamba-1.5-Large | 🟢 FREE | 16 | 16 free models; quotas tied to Copilot plan, reset periodically updated 2026-08-24 | https://models.github.ai/inference |
| ModelScope MiniMax/MiniMax-M2.5 (204K) · qwen-qwen3-5-35b-a3b · qwen-qwen3-5-27b | 🟢 FREE | 58 | 58 free models (largest catalog); 2K RPD shared pool, ≤500/model updated 2026-08-24 | https://api-inference.modelscope.cn/v1 |
| OVHcloud AI Endpoints | 🟢 FREE | 14 | 14 free models; 2 RPM anonymous updated 2026-08-24 | https://oai.endpoints.kepler.ai.cloud.ovh.net/v1 |
| Cloudflare Workers AI @cf/meta/llama-3.3-70b-instruct-fp8-fast · @cf/qwen/qwen1.5-7b-chat · @cf/mistral/mistral-7b-instruct-v0.1 | 🟢 FREE | 40 | 40 free models; 10K neurons/day shared pool updated 2026-08-24 | https://api.cloudflare.com/client/v4/accounts/{account_id}/ai/run |
| Z AI (Zhipu) glm-5.2 (1M) · glm-5.1 (202K) | 🟢 FREE | 4 | 4 free GLM-family models, permanent updated 2026-08-24 | https://open.bigmodel.cn/api/paas/v4 |
| SambaNova | 🟢 FREE | 4 | 4 free models, permanent updated 2026-08-24 | https://api.sambanova.ai/v1 |
| Cohere command-a-218b (436K) · command-a-111b (288K) · command-r | 🟢 FREE | 12 | 12 free models on trial key; 20 RPM updated 2026-08-24 | https://api.cohere.com/v2 |
| Hugging Face Router router auto-select | 🟢 FREE | 7 | 7 free models; 100K credits/month renewable updated 2026-08-24 | https://router.huggingface.co/v1 |
| SiliconFlow | 🟢 FREE | 3 | 3 free models, permanent updated 2026-08-24 | https://api.siliconflow.cn/v1 |
| LLM7.io DeepSeek-V4-Flash-0731 · Inkling · XiaomiMiMo/MiMo-V2.5 | 🟢 FREE | 16 | 16 free models; 30 RPM (120 with token) probe: live · 45 models · 2026-09-11 updated 2026-09-11 | https://api.llm7.io/v1 |
| Kilo Code | 🟢 FREE | 12 | 12 free models; ~200 req/hr quota updated 2026-08-24 | — |
| OpenCode Zen | 🟢 FREE | 12 | 12 free models, permanent updated 2026-08-24 | https://opencode.ai/zen/v1 |
| Aion Labs | 🟢 FREE | 7 | 7 free models, permanent updated 2026-08-24 | https://api.aionlabs.ai/v1 |
| Agnes AI | 🟢 FREE | 5 | 5 free models, permanent updated 2026-08-24 | https://apihub.agnes-ai.com/v1 |
| Chutes.ai | 🟢 FREE | 2 | 2 free models, permanent updated 2026-08-24 | https://api.chutes.ai/v1 |
| Glhf.chat | 🟢 FREE | 2 | 2 free models, permanent updated 2026-08-24 | https://glhf.chat/api/openai/v1 |
| Grok (xAI) | 🟡 CHANGED | 2 | Reported 2 free models; xAI-direct check showed 0 online (2026-08-24) updated 2026-08-24 | https://api.x.ai/v1 |
| Cline | 🟢 FREE | 3 | 3 free models, permanent updated 2026-08-24 | — |
| Alibaba Cloud Model Studio | ⚠️ AT RISK | 5 (0 online) | Reported 5 free models, 0 currently online updated 2026-08-24 | https://dashscope-intl.aliyuncs.com/compatible-mode/v1 |
| DeepSeek | ⚠️ AT RISK | 2 (0 online) | Reported 2 free models, 0 currently online updated 2026-08-24 | https://api.deepseek.com/v1 |
| Nscale | ⚠️ AT RISK | 2 (0 online) | Reported 2 free models, 0 currently online updated 2026-08-24 | https://inference.api.nscale.com/v1 |
| Nebius | ⚠️ AT RISK | 1 (0 online) | Reported 1 free model, 0 currently online updated 2026-08-24 | https://api.studio.nebius.com/v1 |
| AI21 Labs | ⚠️ AT RISK | 2 (0 online) | Reported 2 free models, 0 currently online updated 2026-08-24 | https://api.ai21.com/studio/v1 |
Quick picks
Start here, not at 30 signups.
Best for coding agents
Poolside Laguna S/XS 2.1 (NIM + OpenRouter) · Cohere north-mini-code · Groq groq/compound
Best for apps — no daily cap
NVIDIA NIM — 40 RPM, permanent, no daily request cap
Least signup friction
Groq · OpenRouter · Cloudflare · Cerebras — email-only, no credit card
Free embeddings / rerank
NVIDIA NIM — nemotron-3-embed-1b · llama-nemotron-rerank-vl-1b-v2
Best for China-hosted models
ModelScope (58 free) · SiliconFlow · Z.ai — with data-residency caveats
OpenRouter promo collection
37 models at 10–90% off, one key.
Cheapest available provider per model, via a single OpenRouter integration. Synced from OpenRouter's discounted collection every 8 hours — curated extras noted where the collection lags.
| Model | Discount | Context | Input /M | Output /M |
|---|---|---|---|---|
| OpenAI GPT Latest | 50% | 1.05M | $2 | $10 |
| OpenAI GPT Sol Latest | 50% | 1.05M | $2 | $10 |
| OpenAI: GPT-5.6 Sol | 50% | 1.05M | $2 | $10 |
| OpenAI: GPT-5.6 Sol (batch) | 50% | 1.05M | $1 | $5 |
| OpenAI: GPT-5.6 Sol Pro | 50% | 1.05M | $2 | $10 |
| OpenAI: GPT-5.6 Sol Pro (batch) | 50% | 1.05M | $1 | $5 |
| Inception: Mercury 2.5 | 80% | 260K | $0.04 | $0.15 |
| Qwen: Qwen3 235B A22B Instruct 2507 | 75% | 262K | $0.0875 | $0.35 |
| Upstage: Solar Pro 4 | 70% | 524K | $0.09 | $0.36 |
| Z.ai: GLM 5.2 | 65% | 1.05M | $0.4872 | $1.531 |
| inclusionAI: Ling 3.0 Flash | 65% | 262K | $0.021 | $0.063 |
| Meituan: LongCat 2.0 | 60% | 1.05M | $0.30 | $1.20 |
| Qwen: Qwen3 30B A3B Instruct 2507 | 55% | 262K | $0.04815 | $0.1931 |
| DeepSeek: DeepSeek V4 Flash Vision Exp | 51% | 1.05M | $0.2156 | $0.6468 |
| Google Gemini Flash Latest | 50% | 1.05M | $0.75 | $3.75 |
| Z.ai: GLM Flash Latest | 50% | 1.31M | $0.075 | $0.25 |
| Google: Gemini 3.7 Flash | 50% | 1.05M | $0.75 | $3.75 |
| Google: Gemini 3.7 Flash (batch) | 50% | 1.05M | $0.375 | $1.875 |
| Google: Gemini 3.8 Flash | 50% | 1.05M | $0.75 | $3.75 |
| Google: Gemini 3.8 Flash (batch) | 50% | 1.05M | $0.375 | $1.875 |
| Tencent: Hy3 | 50% | 262K | $0.07 | $0.29 |
| Z.ai: GLM 5.3 Flash | 50% | 1.31M | $0.075 | $0.25 |
| Poolside: Laguna XS 2.1 | 40% | 262K | $0.06 | $0.12 |
| Z.ai: GLM 5 | 40% | 205K | $0.60 | $1.92 |
| MoonshotAI: Kimi K2.6 | 39% | 262K | $0.5795 | $2.44 |
| Z.ai: GLM 5.1 | 31% | 205K | $0.9646 | $3.032 |
| MiniMax: MiniMax M2.7 | 30% | 205K | $0.21 | $0.84 |
| Xiaomi: MiMo-V2.5-Pro | 30% | 1.05M | $0.3045 | $0.609 |
| DeepSeek: DeepSeek V3.2 | 28% | 164K | $0.2088 | $0.3096 |
| StepFun: Step 3.7 Flash | 20% | 262K | $0.16 | $0.92 |
| Alibaba: Wan 3.0 | 15% | - | from $0.0425/sec | - |
| MiniMax: MiniMax M2 | 15% | 205K | $0.255 | $1.02 |
| Xiaomi: MiMo-V2.5 | 15% | 1.05M | $0.119 | $0.238 |
| DeepSeek: DeepSeek V3 | 10% | 164K | $0.2574 | $1.029 |
| Poolside: Laguna S 2.1 | 10% | 1.05M | $0.09 | $0.18 |
| Z.ai: GLM Latest | 3% | 1.31M | $0.97 | $3.308 |
| Z.ai: GLM 5.3 | 3% | 1.31M | $0.97 | $3.308 |
Beyond free
Cheap providers & startup credits.
Cheapest LLM APIs
Hypereal AIPremium models (Claude/GPT/Gemini) + media; free tier 60 RPM
Blackmagic AIPrepaid multi-provider gateway, 48–74% off list
DeepSeekFrontier-class reasoning on a budget + off-peak discounts
Gemini 3.5 FlashLowest big-name flash tier for high volume
GroqFast + cheap open models, low rate high speed
DeepInfraLowest open-model per-token hosting
Together AICompetitive open rates + fine-tuning
Fireworks AIProduction open models + structured output
Cloud credit programs
AWS Activate$1K–$300K · AWS-native stacks; $1K–$8.5K bootstrap tier
Google for Startups$2K–$350K · AI-first; $350K AI tier
Microsoft for Startups$1K–$150K · Easiest entry; Azure OpenAI + GitHub
Cloudflare for Startups$5K–$250K · Workers AI, R2, CDN
Oracle for Startups$500 + 70% ongoing · OCI workloads; most accessible
IBM Cloud for Startupsup to $120K/yr · Watson AI
DigitalOcean HatchCustom · AI/ML priority; MVPs
OVHcloud Startup€10K–€100K · EU infra, no equity
Vercel for Startups~$2.4K + AI Accelerator · Frontend/Edge
Now put this list behind one endpoint.
ModelPanel turns this index into routing — the gateway picks a live model from the free tier, falls back when one dies, and annotates who served. Snapshot 2026-09-11.
How ModelPanel uses this