CYBIRD-D/FREE-LLM-API-Provider
CYBIRD-D/FREE-LLM-API-Provider has 123 GitHub stars and 16 forks. GitHot measured 0 stars in 24 hours and +2 in 7 days. Last checked 2026-11-15 15:44 UTC. Last push 2026-11-13.
Measured star growth
- in 24h
- 0
- in 7 days
- +2
- in 15 days
- Update delayed
- in 1 month
- Update delayed
Net star changes over overlapping periods, not daily totals. Missing measurements are left blank. Stars measure attention, not software quality.
About
This is a list of free LLM provider resources accessible via API. Including some cn platform 免费llm api平台(包括cn平台)
AIREADME.md
OriginalFree LLM API Provider
English | 简体中文 | 日本語 | Português (Brasil)
Last reviewed: 2026-11-02
This directory tracks free LLM inference APIs, their account requirements and rate limits (including CN providers).
Free inference, one-time trial credits and $0/month subscriptions with paid token usage are not the same thing.
Free quotas and model catalogs can change.
免费 LLM API 平台汇总(包括中国大陆平台)。免费调用、一次性试用额度和按量付费(免月费)应分别判断。
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
You may also want to read my other posts:
- How to Choose Your LLM For Translation?
- Local LLMs Collection For Translation
- LLM Timeline
- LLM API Official Pricing Tracker
Easy guide to deploy online LLM api(luna):
- Sign up/register in certain platform;
- Get this platform's Api Key and Endpoint Address
- Check if there's an available date for the key
- Put it in the softwares that support it.
Content/目录
- Global Platform
- 中国大陆平台
- LLM PRICE LIST
Global Platform
AMD Radeon Cloud Token Factory (China site)
Last checked: 2026-11-02
Free shared-model API with a daily renewable usage cap
Official links: Token Factory console · Documentation · Rate limits · Usage and quota · Terms
API base URL (OpenAI-compatible):
https://developer.amd.com.cn/radeon/api/v1. Also supports Anthropic-compatible Messages requests. No GPU instance needed.Free quota: Resets daily at 00:00 Beijing time (UTC+8). Usage is metered in pts (points), not USD, GPU cloud credits, or a fixed number of tokens.
- Daily pts limits vary by account and can change; the official 10-pt example is not a guaranteed allowance. (Usually 1 pt)
- Daily pts limits vary by account and can change; the official 10-pt example is not a guaranteed allowance. (Usually 1 pt)
Typical rate limits (not guaranteed): 20 RPM/account at the gateway, 30 RPM/API key at admission, and 8 concurrent requests/API key
Models&Price(expand to see)
| Model | Input | Output | 1 pt's Tokens(input:output = 1:1) |
|---|---|---|---|
| DeepSeek-V4-Flash | 0.14 | 0.28 | 4.76M |
| DeepSeek-V4-Flash-Vision-Exp | 0.14 | 0.28 | 4.76M |
| DeepSeek-V4.1-Flash | 0.28 | 0.56 | 2.38M |
| Qwen3.8-Flash-Next | 0.14 | 0.28 | 4.76M |
| Qwen3.8-27B | 0.035 | 0.07 | 19.05M |
| GLM-5.3-Flash | 0.28 | 0.56 | 2.38M |
| MiMo-V2.6-Flash | 0.14 | 0.28 | 4.76M |
| MiniCPM5-2B | 0.035 | 0.07 | 19.05M |
| MiniCPM-V-4.7-1B | 0.035 | 0.07 | 19.05M |
| MiniCPM-V-4.7-35B-A3B | 0.035 | 0.07 | 19.05M |
| Gemma-4-31B-IT | 0.035 | 0.0 | 19.05M |
AMD AI Developer Program (one-time cloud credits)
Last checked: 2026-11-02
Promotional credits, not a recurring free LLM API tier
- Official links: Join AMD AI Developer Program · Benefits, FAQ and credit terms
- Membership is free for approved developers. After joining, apply through member perks and select either of the following benefits (do not assume both can be claimed):
- AMD Developer Cloud — US$100 credit for AMD cloud GPU compute (which can be used to deploy or run models). The credits expire 30 days after activation via the redemption link.
- Fireworks AI — US$50 credit on the third-party Fireworks platform (including eligible model inference). The redemption code must be used within 90 days of issuance; after redemption, the credits remain available until consumed, subject to Fireworks' terms.
- Important: AMD Developer Cloud credit is GPU infrastructure credit, not a ready-made hosted LLM API allowance. Both options are one-time promotions, not daily-renewable free API quotas; approval and eligibility terms apply. Do not confuse this global program with the China developer program's separate Devpoint-to-GPU-hours benefit.
Google Gemini / Gemma 4
Last checked: 2026-10-30
- Free-tier model pricing · Rate limits · Your project's limits in AI Studio
- Endpoint:
https://generativelanguage.googleapis.com - Free-tier examples confirmed on Google's pricing page:
gemini-3.8-flash,gemini-3.5-flash-liteand Gemma 4. Check the Free Tier column for other models. - Not free-tier eligible:
gemini-4-argon,gemini-3.1-pro-preview(other Pro models require individual checks). - Google does not publish a single guaranteed RPM / TPM / RPD quota that applies to every free project. Limits depend on model, project and account tier and are shown in AI Studio. Previously observed numbers (e.g. 5 RPM/20 RPD) are not universal guarantees.
- Limits apply per project (not per key); RPD quotas reset at midnight Pacific time. Free-tier prompts may be used to improve Google products.
- ★Alternative, unofficial: AI Studio to API Adapter. Review its security and terms before use.
| Model | Requests/minute (RPM) | Tokens/minute (TPM) | Requests/day (RPD) |
|---|---|---|---|
| Gemini 2.5~3.8 Flash | 5 | 250k | 20 |
| Gemini 2.5 Flash Lite | 10 | 250k | 20 |
| Gemini 3.1/3.5 Flash Lite | 15 | 250k | 500 |
| Gemma 4 26B/31B | 30 | 16k | 14.4K |
| Gemma 3 (1B/2B/4B/12B/27B) | 30 | 15k | 14.4k |
| gemini 2 Flash/Lite | |||
| gemini 2.5/3.1 Pro |
Nvidia
Last checked: 2026-10-30
https://build.nvidia.com/models
Endpoint (OpenAI-compatible): https://integrate.api.nvidia.com/v1
Rate limit: Usually up to 40 Requests per minute(RPM)
Nvidia: Maximum API requests accepted in a given timeframe.
Rate limits may vary by model and traffic from other users may cause throttling.
For dedicated availability, deploy models as a dedicated endpoint with NVIDIA NIM.
- Representative catalog below, not a fixed or exhaustive 55-model list. Verify the Free Endpoint in the live NVIDIA catalog.
- OpenAI/Google/Meta/Microsoft/NVIDIA/Mistral AI/DeepSeek/Moonshot AI/Z.ai/MiniMax etc.
- Free endpoints are for trial/evaluation, check NVIDIA trial privacy terms before submitting sensitive data.
Model List
DeepSeek
deepseek-ai/deepseek-v4-flash
deepseek-ai/deepseek-v4-flash-0731
deepseek-ai/deepseek-v4-pro
deepseek-ai/deepseek-v4.1-flash
google/codegemma-7b
google/gemma-7b
Meta
meta/llama2-70b
meta/llama-3.1-8b-instruct
meta/llama-3.1-70b-instruct
meta/llama-3.2-1b-instruct
meta/llama-3.2-3b-instruct
meta/llama-3.3-70b-instruct
meta/muse-glimmer-30b
Microsoft
microsoft/phi-4-mini-instruct
microsoft/phi-4-mini-flash-reasoning
MiniMax
minimaxai/minimax-m2.5
minimaxai/minimax-m2.7
minimaxai/minimax-m3.1-flash
Mistral
mistralai/mistral-nemotron
mistralai/mixtral-8x7b-instruct
mistralai/mixtral-8x22b-instruct
mistralai/mistral-large-4-1t-a52b
Moonshot / Kimi
moonshotai/kimi-k2-instruct
moonshotai/kimi-k2-thinking
moonshotai/kimi-k3
NVIDIA
nvidia/gliner-pii
nvidia/llama-3.1-nemoguard-8b-content-safety
nvidia/llama-3.1-nemoguard-8b-topic-control
nvidia/llama-3.1-nemotron-safety-guard-8b-v3
nvidia/llama-3.3-nemotron-super-49b-v1
nvidia/llama-3.3-nemotron-super-49b-v1.5
nvidia/llama-3.1-nemotron-ultra-253b-v1
nvidia/nemotron-3-ultra-550b-a55b
nvidia/nemotron-3.5-lightning-30b-a3b
nvidia/nemotron-3.5-super-vl-120b-a12b
nvidia/nemoguard-jailbreak-detect
nvidia/nemotron-3-nano-30b-a3b
nvidia/nemotron-3-super-120b-a12b
nvidia/nemotron-content-safety-reasoning-4b
nvidia/nvidia-nemotron-nano-9b-v2
nvidia/riva-translate-4b-instruct-v1.1
nvidia/riva-translate-4b-instruct-v2
nvidia/usdcode
OpenAI
openai/gpt-oss-20b
openai/gpt-oss-120b
Poolside
poolside/laguna-xs-2-1
Qwen
qwen/qwen2.5-coder-32b-instruct
qwen/qwen3-next-80b-a3b-instruct
qwen/qwen3-next-80b-a3b-thinking
qwen/qwq-32b
Sarvam AI
sarvamai/sarvam-m
StepFun
stepfun-ai/step-3.5-flash
stepfun-ai/step-5-preview
Stockmark
stockmark/stockmark-2-100b-instruct
Thinking Machines
thinking-machines/inkling
Upstage
upstage/solar-10.7b-instruct
Z.ai / GLM
z-ai/glm4.7
z-ai/glm5.1
z-ai/glm-5.2
z-ai/glm-5.3
z-ai/glm-5.3-flash
Ollama
Last checked: 2026-10-30
- Official pricing and cloud plans · Cloud models · 2026-08-31 pricing transition
- Pay as you go ($0/month) means no subscription, not free inference: cloud token usage consumes purchased credits. The old blanket list of “free models” and 3-hour session resets no longer describes the current published plans.
- Running local models on your own hardware incurs no Ollama cloud token charges.
- Cloud concurrency: 3 for Pay as you go and Pro; 10 for Max and Team. Some legacy accounts may have different conditions.
Free model list
- Gemma4: 31b
- gpt-oss 20b/120b
- nemotron-3-nano 30B/super 120B-A12B/ultra 550B-A55B
- Reflection Beam 501B (limit time)
Session usage reset in 3hours
| Plan | Subscription | Included cloud usage |
|---|---|---|
| Pay as you go | $0/month | None advertised; purchase usage credits |
| Pro | $20/month ($200/year) | $60/month |
| Max | $100/month | $300/month |
| Team | $500/month | $1,000/month shared |
| Enterprise | Custom | Custom |
Current cloud model pricing (USD per 1M input / 1M output tokens)
| Model | Input | Output | Combined |
|---|---|---|---|
| deepseek-v4.1-flash (off-peak) | $0.15 | $0.60 | $0.75 |
| deepseek-v4.1-flash (standard) | $0.30 | $1.20 | $1.50 |
| deepseek-v4-pro (off-peak) | $0.66 | $1.98 | $2.64 |
| deepseek-v4-pro (standard) | $1.32 | $3.96 | $5.28 |
| gemma4 | $0.14 | $0.40 | $0.54 |
| glm-5.3 / glm-5.2 | $1.40 | $4.40 | $5.80 |
| glm-5.3-flash | $0.15 | $0.50 | $0.65 |
| gpt-oss:120b | $0.15 | $0.60 | $0.75 |
| gpt-oss:20b | $0.07 | $0.30 | $0.37 |
| kimi-k3 | $3.00 | $15.00 | $18.00 |
| kimi-k2.7-code / kimi-k2.6 | $0.95 | $4.00 | $4.95 |
| minimax-m3.1-flash | $0.15 | $0.60 | $0.75 |
| minimax-m3 | $0.60 | $2.40 | $3.00 |
| minimax-m2.7 | $0.30 | $1.20 | $1.50 |
| mistral-large-4 | $0.70 | $2.10 | $2.80 |
| mistral-large-3 | $0.50 | $1.50 | $2.00 |
| muse-spark-open | $1.25 | $4.25 | $5.50 |
| nemotron-3-nano | $0.06 | $0.24 | $0.30 |
| nemotron-3-super | $0.015 | $0.60 | $0.615 |
| nemotron-3-ultra | $0.10 | $3.00 | $3.10 |
Off-peak pricing applies to eligible models outside 12:00–18:00 UTC on weekdays, and throughout weekends. See the official table for cached-input prices. This table is paid cloud pricing, not a free-tier model listing.
Groq
Last checked: 2026-10-30
The published free rate-limits table includes GPT-OSS 20B/120B and Qwen3.8 27B. Your account may have different limits.
https://console.groq.com/docs/rate-limits
Endpoint (OpenAI-compatible): https://api.groq.com/openai/v1
Model list
| Model | Request/Minute | Request/Day | Token/Minute | Token/Day |
|---|---|---|---|---|
| openai/gpt-oss-120b | 30 | 1K | 8K | 200K |
| openai/gpt-oss-20b | 30 | 1K | 8K | 200K |
| qwen/qwen3.8-27b | 30 | 1K | 8K | 200K |
Cerebras
> Last checked: 2026-10-30
Unusable rate limit 5$ per account
https://inference-docs.cerebras.ai/support/rate-limits
Endpoint: https://api.cerebras.ai
| Model | Requests/Minute | Tokens/Minute | Tokens/Hour | Tokens/Day |
|---|---|---|---|---|
| gpt-oss-120b | 5 | 30k | 1M | 1M |
| gemma-4-31b | 5 | 30k | 1M | 1M |
OpenRouter
Last checked: 2026-10-30
https://openrouter.ai/collections/free-models
https://openrouter.ai/pricing
Endpoint (OpenAI-compatible): https://openrouter.ai/api/v1
Models: Check the live collection. openrouter/free can route to a free model; $0-priced models without :free can follow different eligibility rules.
- Free usage limits: If you’re using a free model variant (with an ID ending in
:free/(free))
you can make up to20requests/minute.- If you have purchased less than
$10 credits, you’re limited to50freemodel requests/Day. - If you purchase at least
$10 credits, the daily limit increases to1000free model requests/day. Failed requests also count. See official limits for purchase thresholds.
- If you have purchased less than
| Model variant (ID) | Credits purchased | Rate limit (requests/min) | Daily limit (requests/day) |
|---|---|---|---|
| *:free | < $10 credits | 20 | 50 |
| *:free | ≥ $10 credits | 20 | 1000 |
Cloudflare
Official pricing page last updated: 2026-10-20
Last Check: 2026-09-23
https://developers.cloudflare.com/workers-ai/platform/pricing/#llm-model-pricing
https://developers.cloudflare.com/workers/platform/pricing/
- Workers Free: 10,000 Neurons/day (resets 00:00 UTC); paid overage $0.011 per 1,000 Neurons.
"Neurons are our way of measuring AI outputs across different models, representing the GPU compute needed to perform your request. Our serverless model allows you to pay only for what you use without having to worry about renting, managing, or scaling GPUs."
Billing requirement: Certain Kimi, DeepSeek V4/V4.1 and GLM-5.x APIs require Workers Paid or prepaid AI Gateway credits even with unused free Neurons. Listed models are not all available without a billing method.
Models list
- Llama
- Llama2-7b
- Llama3.1-8b/70b
- Llama3.2-1b/3b/11b(vision)
- Llama4-scout-17b-16e
- Qwen
- qwq-32b
- qwen2.5-coder-32b
- qwen3-30b-a3b
- qwen3.8-27b
- Mistral
- Mistral-7b-intruct-v0.1
- Mistral-small-3.1b-24b
- deepseek-r1-distill-qwen-32b
- deepseek V4 flash/pro and V4.1 flash 1017
- Gemma
- gemma-3-12b
- gemma-4-26B-A4B
- gemma-sea-lion-v4-27b-it
- granite-4.0-h-micro
- glm-4.7-flash
- glm-5.2 / glm-5.3 / glm-5.3-flash
- nemotron-3-120b-a12b
- kimi-k2.5/k2.6/k2.7-code
Full list with token cost
Pricing page updated October 20, 2026 (reviewed October 30)
Unit: USD per 1M tokens
Ranking method:
Combined = Input + OutputFree allocation: 10,000 Neurons per day
Free Tokens/day assumes an Input : Output token ratio of
1 : 1.3.Free-token estimates use Neurons-per-token rates rather than rounded USD prices; for billing-gated models these are illustrative, not free account entitlements.
Cached input is excluded from the free-token calculation.
Identical models from the same family are merged only when their pricing and relevant notes are identical.
| Rank | Provider | Model | Input | Output | Combined | 10K Neurons ≈ Tokens/day (I:O = 1:1.3) | Notes |
|---|---|---|---|---|---|---|---|
| 1 | IBM | Granite 4.0 H Micro | $0.017 | $0.112 | $0.129 | ~1.560M | — |
| 2 | Meta | Llama 3.2 1B Instruct | $0.027 | $0.201 | $0.228 | ~878K | — |
| 3 | Mistral | Mistral 7B Instruct v0.1 | $0.110 | $0.190 | $0.300 | ~708K | — |
| 4 | Meta | Llama 3.2 3B Instruct | $0.051 | $0.335 | $0.386 | ~520K | — |
| 5 | Qwen | Qwen3 30B-A3B FP8 | $0.051 | $0.335 | $0.386 | ~520K | — |
| 6 | Meta | Llama 3/3.1 8B Instruct AWQ | $0.123 | $0.266 | $0.389 | ~539K | Identical pricing; merged |
| 7 | Gemma 4 26B-A4B IT | $0.100 | $0.300 | $0.400 | ~516K | — | |
| 8 | Meta | Llama 3.1 8B Instruct FP8 Fast | $0.045 | $0.384 | $0.429 | ~465K | — |
| 9 | Meta | Llama 3.1 8B Instruct FP8 | $0.152 | $0.287 | $0.439 | ~482K | — |
| 10 | Z.AI | GLM-4.7 Flash | $0.060 | $0.400 | $0.460 | ~435K | — |
| 11 | OpenAI | GPT-OSS 20B | $0.200 | $0.300 | $0.500 | ~429K | — |
| 12 | Meta | Llama Guard 3 8B | $0.484 | $0.030 | $0.514 | ~484K | Safety classifier; not a general-purpose chat LLM |
| 13 | Meta | Llama 3.2 11B Vision Instruct | $0.049 | $0.676 | $0.725 | ~273K | Vision-capable |
| 14 | Gemma 3 12B IT | $0.345 | $0.556 | $0.901 | ~237K | — | |
| 15 | AI Singapore | Gemma SEA-LION v4 27B IT | $0.351 | $0.555 | $0.906 | ~236K | — |
| 16 | Mistral | Mistral Small 3.1 24B Instruct | $0.351 | $0.555 | $0.906 | ~236K | — |
| 17 | OpenAI | GPT-OSS 120B | $0.350 | $0.750 | $1.100 | ~191K | — |
| 18 | Meta | Llama 3/3.1 8B Instruct | $0.282 | $0.827 | $1.109 | ~187K | Identical pricing; merged |
| 19 | Meta | Llama 4 Scout 17B-16E Instruct | $0.270 | $0.850 | $1.120 | ~184K | — |
| 20 | Qwen | QwQ 32B | $0.660 | $1.000 | $1.660 | ~129K | — |
| 21 | Qwen | Qwen2.5 Coder 32B Instruct | $0.660 | $1.000 | $1.660 | ~129K | — |
| 22 | DeepSeek | DeepSeek V4 Flash 0731 | $0.440 | $1.320 | $1.760 | ~117K | Cached input $0.014; paid billing method required |
| 23 | NVIDIA | Nemotron 3 120B-A12B | $0.500 | $1.500 | $2.000 | ~103K | — |
| 24 | Meta | Llama 3.1/3.3 70B Instruct FP8 Fast | $0.293 | $2.253 | $2.546 | ~78.5K | Identical pricing; merged |
| 25 | Moonshot AI | Kimi K2.5 | $0.600 | $3.000 | $3.600 | ~56.2K | Cached input $0.100 |
| 26 | Qwen | Qwen3.8 27B | $0.450 | $3.200 | $3.650 | ~54.9K | — |
| 27 | Moonshot AI | Kimi K2.6 | $0.950 | $4.000 | $4.950 | ~41.1K | Cached input $0.160; paid billing method required |
| 28 | Moonshot AI | Kimi K2.7 Code | $0.950 | $4.000 | $4.950 | ~41.1K | Cached input $0.190; paid billing method required |
| 29 | DeepSeek | DeepSeek V4 Pro 0813 | $1.320 | $3.960 | $5.280 | ~39.1K | Cached input $0.044; paid billing method required |
| 30 | DeepSeek | DeepSeek R1 Distill Qwen 32B | $0.497 | $4.881 | $5.378 | ~37.0K | — |
| 31 | Z.AI | GLM-5.2 | $1.400 | $4.400 | $5.800 | ~35.5K | Cached input $0.260; paid billing method required |
| 32 | Meta | Llama 2 7B Chat FP16 | $0.556 | $6.667 | $7.223 | ~27.4K | — |
Cohere
Last web updated: 2026-06-09
Last Check: 2026-10-30
https://docs.cohere.com/docs/rate-limits
Endpoint: https://api.cohere.ai/compatibility
| Endpoint | Trial rate limit (requests/min) | Trial monthly cap (calls/month) |
|---|---|---|
| Chat API | 20/min (per model) | 1000/month |
- Chat model includes
- Command A+
- Command A Reasoning
- Command A Translate
- Command A Vision
- Command A
- Command R+
- Command R
- Command R7B
- North Mini Code
All endpoints are limited to 1,000 calls per month with a trial key
Z.ai (GLM-4.5/4.7-Flash)
Last Check: 2026-10-30
https://docs.z.ai/guides/overview/pricing
Endpoint: https://api.z.ai/api/paas/v4
Free Models :
- GLM-4.5-Flash
- GLM-4.7-Flash
- GLM-4.6V-Flash
No offical rate usage limits
Github
https://docs.github.com/en/github-models/use-github-models/prototyping-with-ai-models#rate-limits
GitHub Models has been retired.
As of July 30, 2026, GitHub Models has been fully retired. The playground, model catalog, inference API, and bring your own key (BYOK) are no longer available to any customer.
GitHub Models was a separate service from GitHub Copilot and is unrelated to GitHub Copilot services.
Mistral
Last checked: 2026-10-30
Official rate-limit guide · Explanation
Free-mode limits depend on the organization and model. The old numeric cap below is a historical community report, not a guaranteed free-tier entitlement.
Check Admin Panel → API → Limits.
Endpoint: https://api.mistral.ai
- From community&reports; No offcial list
| Plan / Tier | Requests/second (RPS) | Tokens/minute (TPM) | Tokens/month |
|---|---|---|---|
| Mistral API Free | 1/sec | 500k | 1 billion |
SambaNova
Last checked: 2026-10-30
https://docs.sambanova.ai/docs/en/models/rate-limits#free-tier
| Model | Requests/Minute | Requests/Day | Tokens/Day |
|---|---|---|---|
| DeepSeek-V3.1 | 20 | 20 | 200k |
| Meta-Llama-3.3-70B-Instruct | 20 | 20 | 200k |
| gpt-oss-120b | 20 | 20 | 200k |
| DeepSeek-V3.2 | 20 | 20 | 200k |
| gemma-4-31B-it | 20 | 20 | 200k |
AionLabs
Last checked: 2026-10-30
https://www.aionlabs.ai/docs/rate-limits/
https://www.aionlabs.ai/docs/models/
Endpoint: https://api.aionlabs.ai/v1
Shared free-tier caps: 15 RPM, 20K TPM, 20K tokens/day (official rate limits).
| Model | Requests/Minute | Tokens/Day |
|---|---|---|
| Aion-2.0 (DeepSeek V3.2 finetune) | 15 | 20k |
| Aion-2.5 (Improve of Aion-2.0) | 15 | 20k |
| Aion-3.0 (GLM) | 15 | 20k |
| Aion-3.0 Mini (Deepseek) | 15 | 20k |
| Aion-3.5 (GLM) | 15 | 20k |
| Aion-3.5 Mini (GLM) | 15 | 20k |
| Aion-RP 1.0 (8B) (llama-3.1-8b fineune/openweight) | 15 | 20k |
SKT
Free Api for A.X 4.0 (7B/72B, based on Qwen2.5)
Korean⇌EN
https://github.com/SKT-AI/A.X-4.0/blob/main/apis/README.md
IBM
Last Check: 2026-10-08
https://www.ibm.com/products/watsonx-ai/pricing
| Plan / Tier | Requests/second (RPS) | Tokens/month |
|---|---|---|
| watsonx.ai Free tier (Foundation Models) |
2/sec | 300k |
Scaleway (1M free token/per account-no refresh)
https://www.scaleway.com/en/docs/generative-apis/faq/#how-does-the-free-tier-work
Together AI (certain free-endpoint:no text models now)
> Last updated: 2025-11-17
https://www.together.ai/models
- Llama3-70b
https://www.together.ai/models/llama-3-3-70b-free
https://www.together.ai/models/deepseek-r1-distilled-llama-70b-free
NO FREE TIER Now
https://support.together.ai/articles/1862638756-changes-to-free-tier-and-billing-july-2025
CN Platform
AMD Radeon Cloud Token Factory(AMD 免费模型 API)
免费方式: 每日重置的共享模型推理 API 额度,以 pts 计量;不是一次性体验金,也不是 GPU 算力券。
官方资料: Token Factory 控制台 · 中文使用文档 · 用量及额度 · 限流规则 · API 使用条款
API 地址(兼容 OpenAI): https://developer.amd.com.cn/radeon/api/v1;亦支持兼容 Anthropic Messages 的接口。
- 每天多少额度: 每日北京时间 00:00(UTC+8) 重置。不同账号的 pts/天额度可能不同,也可能调整;
- 官方文档中的 10 pts/天只是示例,不能当作统一赠送额度。(一般为1pt)
- pts 代表什么: 是模型调用的用量计算单位,不是现金、开发者积分、GPU 算力券,也不等于固定数量的 tokens;具体模型按每百万输入/输出 tokens 等分别计量,以模型卡显示的 pts 单价为准。
- 限流参考: 网关通常 20 RPM/账号,接入层通常 30 RPM/API Key,每个 API Key 通常最多 8 个并发请求;实际还受模型后端共享容量限制。
模型&价格(展开显示完整表格)
| 模型 | 输入价格 | 输出价格 | 1 pt 可用总 Tokens(输入:输出 = 1:1) |
|---|---|---|---|
| DeepSeek-V4-Flash | 0.14 | 0.28 | 4.76M |
| DeepSeek-V4-Flash-Vision-Exp | 0.14 | 0.28 | 4.76M |
| DeepSeek-V4.1-Flash | 0.28 | 0.56 | 2.38M |
| Qwen3.8-Flash-Next | 0.14 | 0.28 | 4.76M |
| Qwen3.8-27B | 0.035 | 0.07 | 19.05M |
| GLM-5.3-Flash | 0.28 | 0.56 | 2.38M |
| MiMo-V2.6-Flash | 0.14 | 0.28 | 4.76M |
| MiniCPM5-2B | 0.035 | 0.07 | 19.05M |
| MiniCPM-V-4.7-1B | 0.035 | 0.07 | 19.05M |
| MiniCPM-V-4.7-35B-A3B | 0.035 | 0.07 | 19.05M |
| Gemma-4-31B-IT | 0.035 | 0.07 | 19.05M |
最后核查:2026 年 10 月 30 日
本节区分持续免费推理、定期发放的额度、一次性试用 Token/代金券和非模型 API 的产品积分。免费模型、使用限制、活动及地区政策可能随时变化,请以登录后的平台控制台为准。本次核查依据平台文档及公告,未使用实际账号逐一测试 API。
魔搭社区(ModelScope)
免费方式: 每日魔粒积分,可用于模型推理;并非无限次免费 API。
官方资料: API-Inference 调用限制 · 魔粒体系公告(2026-08-10) · 魔粒使用说明
API 地址(兼容 OpenAI): https://api-inference.modelscope.cn/v1
- 使用 API-Inference 需要绑定已完成实名认证的阿里云账号。
- 根据现行魔粒体系,每日登录可领取 200 魔粒,绑定阿里云账号可额外获得 50 魔粒/天;参与社区活动还可能获得额外积分。
- 模型推理消耗参考:轻量模型约 0.5 魔粒/次,主流模型约 1 魔粒/次,旗舰模型约 2 魔粒/次。实际以具体模型的计费规则为准。
- 旧版规则列有所有模型合计 2,000 次/天、单模型最多 500 次/天,部分大型模型限制更严格。这些数字不能直接当作当前账号每天可获得的免费调用次数;还需考虑魔粒余额、模型限额及并发限制。
- 免费共享推理资源主要用于体验和开发测试,不保证生产环境服务等级;使用前请查看账号剩余积分及模型限制。
硅基流动(SiliconFlow)
免费方式: 部分模型长期标价为 0 元;另有活动代金券,两者应分开计算。
官方资料: 免费模型及限流规则 · 免费与 Pro 模型说明 · 2026 年代金券活动公告
API 地址(兼容 OpenAI): https://api.siliconflow.cn/v1
- 完成实名认证后,可调用模型列表中当前明确标注为免费的模型,相关推理费用为 0 元。
- 免费模型有按模型区分的调用速率限制。部分付费版本带有
Pro/前缀,但不能仅凭模型大小或是否带有该前缀判断是否免费,应以模型详情页的实际价格为准。 - 活动代金券: 官方 2026 年「推荐官」活动(2026-01-15 至 2026-12-31)提供实名认证后领取的 16 元通用代金券,须在活动中心主动领取,领取后 180 天内有效;邀请奖励另行计算。
- 仓库此前所写「中国手机号新用户注册即送 2,000 万 tokens」未能确认仍为当前普遍适用的注册权益,因此不再作为确定额度展示。
腾讯云 TokenHub
免费方式: 新用户一次性试用 Token;不属于持续免费的模型 API。
官方资料: TokenHub 中国站新用户免费体验 · TokenHub API 快速入门 · 旧混元平台停服公告
- 腾讯混元旧 API 平台已停止服务: 旧平台于 2026-06-30 停止新购和创建密钥,于 2026-09-30 关闭原控制台及 API 服务。新的模型接入应使用 TokenHub,其提供混元及第三方模型。
- 中国大陆站: 官方新用户活动说明列出语言模型 100 万 tokens 的试用额度,领取后一年内有效;活动公布的结束日期为 2026-12-31。这属于一次性试用权益,部分模型或账号可能有单独限制。
- 需要完成腾讯云实名认证。免费资源包可通过模型列表或开通页面领取,部分权益可能在首次调用时自动发放;最终以控制台显示为准。
- 避免意外扣费: 如不希望试用额度耗尽后产生费用,请关闭后付费模式。未开启后付费时,免费额度用尽后调用将停止。
- 中国站与国际站不可混用: 腾讯云国际站 TokenHub 的试用有效期为 90 天,部分模型开通还可能要求 1 美元余额预授权,与中国站的规则不同。
- 此前的旧版「Hunyuan-lite 免费」和旧混元平台 100 万 tokens / 一年资源包属于历史权益,不能作为已经停服的旧 API 仍可使用的依据。
火山方舟(Volcengine Ark)
免费方式: 模型推理 API 一次性体验额度;与扣子的积分体系不同。
官方资料: 安心体验模式 · 模型开通与免费额度管理
- 安心体验模式: 符合条件的新用户可通过该模式使用 50 万 tokens 模型推理体验额度;额度用尽后,符合条件的推理服务会暂停而不是继续扣费。
- 该模式须在首次开通模型推理服务时选择。已通过其他模式开通服务的账号可能不再符合条件;关闭安心体验模式后不能重新开启。
- 此模式的免费额度仅覆盖适用范围内的模型推理 Token。工具、插件、知识库等其他功能可能单独计费。
- 平台还可能提供其他按模型、活动或账号发放的免费资源包,具体数量与适用条件请查看控制台。
- 旧版「500 资源点/天」及豆包模型资源点消耗表属于**扣子(Coze)**的计费体系,不是火山方舟 API 的免费额度或定价,因此已从本节移除。
扣子(Coze)
免费方式: AI 应用与智能体平台积分;不等同于通用模型推理 API 免费额度。
官方资料: 2026 年订阅体系升级公告 · 现行订阅套餐 · 积分规则
- 2026-01-18 的套餐升级将「资源点」更名为「积分」。升级公告提出个人免费版每日签到领取 1,500 积分(当天有效),取代此前的 500 资源点/天。
- 现行资料存在口径差异: 当前套餐表未将每月订阅积分列入个人免费版;积分规则又明确列出新用户一次性获得 1,500 积分(30 天有效),相关常见问题仍提及每日签到奖励。因此,不能在未查看账号活动页面的情况下保证所有用户每天都能固定领取 1,500 积分。
- 扣子积分主要用于平台产品和功能,不能直接换算成火山方舟 API 的免费 Token 数量。
- 原仓库中按旧版扣子资源点计算的豆包、DeepSeek、Kimi 模型消耗表已移除,以免把过期计费规则当作现行价格。
快手万擎(StreamLake Vanchin)
免费方式: 新用户或活动提供的模型在线推理试用资源包。
官方资料: 免费推理额度说明(2025-12-12 更新) · 模型计费说明
- 需要完成注册、认证和服务开通,方可使用符合条件的免费推理资源包。
- 免费资源包通常按具体模型发放,同一主账号下共享;未确认所有新用户均能领取相同数量的 Token,因此不填写统一免费额度。实际余额请在费用中心 → 资源包中查看。
- 资源包可抵扣符合条件的在线推理输入、输出及缓存命中 Token,但不能抵扣批量推理。
- 注意超额费用: 免费额度过期或用尽后继续调用,可能产生按量计费账单。建议及时查看用量,并在支持的情况下限制或关闭后付费调用。
讯飞星火(iFLYTEK Spark)
免费方式: 部分模型标价为 0 元,另有独立的活动赠送额度。
官方资料: Spark Lite HTTP API 文档 · Spark-X2.5 模型说明 · 星火模型与价格页面
- Spark Lite: 官方 HTTP API 文档标注支持免费使用。
- Spark Lite 经典 API 地址(兼容 OpenAI):
https://spark-api-open.xf-yun.com/v1/。 - Spark-X2.5-1.7B(模型 ID:
spark-x2.5-1.7b):现行模型列表标价为 0 元/百万 tokens。 - Spark-X2.5-4B(模型 ID:
spark-x2.5-4b):模型列表出现 0 元/百万 tokens / 限时免费标注,不可视为永久免费。X2.5 系列属于星火模型平台,调用前需另行确认适用的 API 地址及鉴权方式。 - 旧版「首次开通后,所有模型共享 20 万免费 tokens,有效期一年」未能找到明确的现行依据,故不再将其列为确定权益。
心流(iFlow,已停止服务)
状态:原免费 LLM API 与托管模型库已于 2026 年 4 月 17 日停止服务,不再属于当前可用的免费模型 API 平台。
官方资料: 2026-03-20 停服公告 · 2026-04-17 停服通知
- 原有 iFlow API 服务及托管模型库于 2026-04-17 关闭,相关 iFlow CLI 内置模型访问也随之停止。
- iFlow CLI 仍可配置其他平台的兼容 API,但这不代表原心流托管模型服务恢复。
- 本条仅作历史记录,不应再列为可用免费 API 推荐项。
LLM PRICE LIST
LLM API Pricing — 2026-10-30
Updated from the LLM API Official Pricing Tracker (source table dated 2026-10-30). This is a synchronized reference, not a live price feed. Provider prices and promotions can change; check the individual Pricing Source before purchasing.
- Unit: USD per 1 million tokens.
- Combined = Input + Output: the price of 1M uncached input tokens plus 1M output tokens (not the cost per 1M total tokens).
- Cached Input:
—means the cache price is unknown / not listed, not zero.- Order: lowest Combined cost first. Ties use the lowest directly comparable confirmed Cached Input price.
- Promotions, regular/future prices and peak/off-peak rates occupy separate rows when present in the source tracker. Faster or priority API variants are also included where the source lists separately priced variants.
- Context tiers: use the default/lowest tier for ranking; higher-tier rates appear in Notes.
- Models: same-family variants are combined only when the upstream tracker treats their prices and conditions as equivalent.
- Qwen: Alibaba Cloud Model Studio Singapore / International pricing.
- Table layout preserved: Rank, Provider, Model, Input, Cached Input, Output, Combined, Notes and Pricing Source. Open/Closed status and parameter counts can be found in the full tracker.
| Rank | Provider | Model | Input | Cached Input | Output | Combined | Notes | Pricing Source |
|---|---|---|---|---|---|---|---|---|
| 1 | Qwen | Qwen3.7 Flash | $0.03 | — | $0.13 | $0.16 | ≤32K; 32K–256K: $0.10/$0.40; 256K–1M: $0.20/$0.80 | Source |
| 2 | Meta / OpenRouter | Muse Spark 1.2/1.3 Contributor | $0.10 | $0.002 | $0.20 | $0.30 | 1M context; prompts/outputs may be used to improve Meta products | Source |
| 3 | Qwen | Qwen3.5 Flash | $0.10 | — | $0.40 | $0.50 | International | Source |
| 4 | Anthropic | Claude Haiku 5.5 | $0.10 | $0.01 | $0.50 | $0.60 | ≤100K; >100K: $0.50 input / $0.05 cache / $2.50 output | Source |
| 5 | OpenAI | GPT-6 Luna | $0.10 | $0.01 | $0.50 | $0.60 | 1.05M context; >272K: $0.20/$0.75 | Source |
| 6 | Qwen | Qwen3.8 Flash | $0.15 | — | $0.47 | $0.62 | International; 1M context | Source |
| 7 | GLM / Z.AI | GLM-5.3-Flash | $0.15 | $0.03 | $0.50 | $0.65 | Standard price | Source |
| 8 | DeepSeek | DeepSeek V4.1 Flash 1017 — Off-peak | $0.15 | $0.003 | $0.60 | $0.75 | 1M context; 1017 update (unchanged API rates); off-peak pricing | Source |
| 9 | MiniMax | MiniMax M3.1 Flash | $0.15 | $0.015 | $0.60 | $0.75 | 1M context; model weights now publicly available; multimodal input | Source |
| 10 | Mistral | Mistral Small 4 | $0.15 | $0.015 | $0.60 | $0.75 | 256K context | Source |
| 11 | OpenAI | GPT-5.6 Luna | $0.20 | $0.02 | $1.20 | $1.40 | >272K: $0.40/$1.80 | Source |
| 12 | Meta / OpenRouter | Muse Glimmer 30B | $0.30 | $0.04 | $1.10 | $1.40 | 131K context | Source |
| 13 | OpenAI | GPT-5.4 nano | $0.20 | $0.02 | $1.25 | $1.45 | — | Source |
| 14 | DeepSeek | DeepSeek V4.1 Flash 1017 — Peak | $0.30 | $0.006 | $1.20 | $1.50 | 1M context; 1017 update (unchanged API rates); peak pricing | Source |
| 15 | MiniMax | MiniMax M2.7 | $0.30 | $0.06 | $1.20 | $1.50 | ~205K context; cache write $0.375/MTok | Source |
| 16 | MiniMax | MiniMax M3 | $0.30 | $0.06 | $1.20 | $1.50 | ≤512K prompt; >512K: $0.60 input / $0.12 cache / $2.40 output; permanent 50% discount | Source |
| 17 | Qwen | Qwen3.7 Plus (Promo) | $0.32 | — | $1.28 | $1.60 | International alias; 20% off; regular $0.40/$1.60 | Source |
| 18 | Gemini 3.1 Flash-Lite | $0.25 | $0.025 | $1.50 | $1.75 | — | Source | |
| 19 | Qwen | Qwen3.6 Flash | $0.25 | — | $1.50 | $1.75 | ≤256K; International | Source |
| 20 | Qwen | Qwen3.7 Plus (Regular Price) | $0.40 | — | $1.60 | $2.00 | ≤256K | Source |
| 21 | MiniMax | MiniMax M3 — Priority | $0.45 | $0.09 | $1.80 | $2.25 | ≤512K prompt; faster priority serving; >512K: $0.90/$0.18/$3.60; permanent 50% discount | Source |
| 22 | DeepSeek | DeepSeek V4 Pro — Off-peak | $0.66 | $0.022 | $1.98 | $2.64 | Off-peak pricing | Source |
| 23 | Gemini 3.5 Flash-Lite | $0.30 | $0.03 | $2.50 | $2.80 | — | Source | |
| 24 | Mistral | Mistral Large 4 | $0.70 | $0.07 | $2.10 | $2.80 | 1M context; Oct 29 open release; current official standard API price | Source |
| 25 | Qwen | Qwen3.5 Plus | $0.40 | — | $2.40 | $2.80 | ≤256K; International | Source |
| 26 | MiniMax | MiniMax M2.7 Highspeed | $0.60 | $0.06 | $2.40 | $3.00 | Same open M2.7 model, faster serving; ~205K context; cache write $0.375/MTok | Source |
| 27 | xAI | Grok Build 0.1 | $1.00 | $0.20 | $2.00 | $3.00 | <200K; ≥200K: $2/$4 | Source |
| 28 | Qwen | Qwen3.6 Plus | $0.50 | — | $3.00 | $3.50 | ≤256K; International | Source |
| 29 | xAI | Grok 4.20/4.3 | $1.25 | $0.20 | $2.50 | $3.75 | <200K; ≥200K: $2.50/$5 | Source |
| 30 | GLM / Z.AI | GLM-5 | $1.00 | $0.20 | $3.20 | $4.20 | — | Source |
| 31 | Gemini 3.6/3.8 Flash (Promo) | $0.75 | $0.075 | $3.75 | $4.50 | Introductory price through Dec 31, 2026; 3.7 retired (redirects to 3.8) | Source | |
| 32 | Kimi | Kimi K2.6 | $0.95 | $0.16 | $4.00 | $4.95 | ~262K context | Source |
| 33 | Kimi | Kimi K2.7 Code | $0.95 | $0.19 | $4.00 | $4.95 | ~262K context | Source |
| 34 | GLM / Z.AI | GLM-5-Turbo | $1.20 | $0.24 | $4.00 | $5.20 | 200K context; agent-tuned Turbo variant | Source |
| 35 | OpenAI | GPT-5.4 mini | $0.75 | $0.075 | $4.50 | $5.25 | — | Source |
| 36 | DeepSeek | DeepSeek V4 Pro — Peak | $1.32 | $0.044 | $3.96 | $5.28 | Peak pricing | Source |
| 37 | Meta / OpenRouter | Muse Spark 1.1/1.2/1.3 | $1.25 | $0.15 | $4.25 | $5.50 | 1M context; standard data policy | Source |
| 38 | Meta / OpenRouter | Muse Spark Open | $1.25 | $0.15 | $4.25 | $5.50 | 1M context; released Oct 15; public weights (safety post-trained, same base as Muse Spark 1.3) | Source |
| 39 | GLM / Z.AI | GLM-5.1 | $1.40 | $0.26 | $4.40 | $5.80 | 200K context | Source |
| 40 | GLM / Z.AI | GLM-5.2/5.3 | $1.40 | $0.26 | $4.40 | $5.80 | 1M context | Source |
| 41 | OpenAI | GPT-6.1 Terra | $1.00 | $0.05 | $5.00 | $6.00 | 1.05M context; >272K prompt: $2 input / $0.10 cache / $7.50 output; 30m cache write $1.25 | Source |
| 42 | xAI | Grok 4.5 | $2.00 | $0.30 | $6.00 | $8.00 | <200K; ≥200K: $4/$12 | Source |
| 43 | xAI | Grok 4.6 | $2.00 | $0.50 | $6.00 | $8.00 | <200K; ≥200K: $4/$12 | Source |
| 44 | xAI | Grok 4.7 | $2.00 | $0.50 | $6.00 | $8.00 | 500K context; ≥200K: $4/$12 | Source |
| 45 | Qwen | Qwen3.8 Max | $2.00 | — | $6.00 | $8.00 | International; 1M context | Source |
| 46 | Gemini 3.6/3.8 Flash (Regular Price) | $1.50 | $0.15 | $7.50 | $9.00 | Standard price from Jan 1, 2027; 3.7 retired (redirects to 3.8) | Source | |
| 47 | Mistral | Mistral Medium 3.5 | $1.50 | $0.15 | $7.50 | $9.00 | 256K context | Source |
| 48 | Kimi | Kimi K2.7 Code Highspeed | $1.90 | $0.38 | $8.00 | $9.90 | ~262K; same K2.7 Code model, faster output | Source |
| 49 | Qwen | Qwen3.7 Max | $2.50 | — | $7.50 | $10.00 | International; 1M context | Source |
| 50 | Anthropic | Claude Sonnet 5.5 | $2.00 | $0.10 | $10.00 | $12.00 | Cache read reduced to $0.10 on Oct 7 | Source |
| 51 | Gemini 4 Argon (Promo) | $2.00 | $0.10 | $10.00 | $12.00 | 1M context; implicit caching only; offer through Nov 30, 2026 | Source | |
| 52 | OpenAI | GPT-6.1 Sol | $2.00 | $0.10 | $10.00 | $12.00 | 1.05M context; >272K: $4/$0.20/$15 | Source |
| 53 | Anthropic | Claude Sonnet 5 | $2.00 | $0.20 | $10.00 | $12.00 | — | Source |
| 54 | OpenAI | GPT-6 Sol | $2.00 | $0.20 | $10.00 | $12.00 | 1.05M context; >272K: $4/$0.40/$15 | Source |
| 55 | Kimi | Kimi K3.1 Pro | $2.00 | $0.20 | $10.00 | $12.00 | 1M context; 5-min cache write $2/MTok; planned Open release by Nov 4 (not yet released) | Source |
| 56 | Gemini 3.1 Pro Preview | $2.00 | $0.20 | $12.00 | $14.00 | ≤200K; >200K: $4/$18 | Source | |
| 57 | OpenAI | GPT-5.6 Terra | $2.00 | $0.20 | $12.00 | $14.00 | >272K: $4/$18 | Source |
| 58 | OpenAI | GPT-5.4 | $2.50 | $0.25 | $15.00 | $17.50 | >272K long-context surcharge | Source |
| 59 | Anthropic | Claude Sonnet 4.6 | $3.00 | $0.30 | $15.00 | $18.00 | — | Source |
| 60 | Kimi | Kimi K3 | $3.00 | $0.30 | $15.00 | $18.00 | 1M context | Source |
| 61 | Anthropic | Claude Opus 5.5 | $4.00 | $0.20 | $20.00 | $24.00 | 1M context | Source |
| 62 | Gemini 4 Argon (Regular Price) | $4.00 | $0.20 | $20.00 | $24.00 | 1M context; implicit caching only; from Dec 1, 2026 | Source | |
| 63 | Kimi | Kimi K3.1 Pro Highspeed | $4.00 | $0.20 | $20.00 | $24.00 | Same K3.1 Pro model, 2–3× faster; 1M context; 5-min cache write $4/MTok | Source |
| 64 | Meta | Muse Wonder | $4.00 | $0.20 | $20.00 | $24.00 | 1M context; Meta Model API / OpenRouter; Wonder Full is subscription-only | Source |
| 65 | OpenAI | GPT-5.6 Sol (Promo) | $4.00 | $0.40 | $20.00 | $24.00 | Promo available at least through Nov 21, 2026; regular $5/$30 | Source |
| 66 | Anthropic | Claude Opus 4.7/4.8/5 | $5.00 | $0.50 | $25.00 | $30.00 | — | Source |
| 67 | OpenAI | GPT-5.5 | $5.00 | $0.50 | $30.00 | $35.00 | — | Source |
| 68 | OpenAI | GPT-5.6 Sol (Regular Price) | $5.00 | $0.50 | $30.00 | $35.00 | Original price; currently $4/$20 promo | Source |
| 69 | Anthropic | Claude Fable 5.1 | $10.00 | $0.25 | $50.00 | $60.00 | 1M context; low 0.025x cache-read multiplier | Source |
| 70 | Anthropic | Claude Fable 5 | $10.00 | $1.00 | $50.00 | $60.00 | — | Source |
| 71 | OpenAI | GPT-6 Astra | $10.00 | $1.00 | $50.00 | $60.00 | 1.05M context; >272K: $20/$2/$75 | Source |
| 72 | OpenAI | GPT-5.4/5.5 Pro | $30.00 | — | $180.00 | $210.00 | Maximum-compute tier | Source |
DeepSeek peak/off-peak: Peak hours are 01:00–04:00 UTC and 06:00–10:00 UTC; all other hours are off-peak. For DeepSeek V4.1 Flash 1017, the source tracker gives $0.15 input / $0.003 cached input / $0.60 output (off-peak) and $0.30 / $0.006 / $1.20 (peak). DeepSeek V4 Pro rates are listed in their own peak and off-peak rows above.











