Free Inference You Can Build With

Free AI & LLM inference you can build with: an API key, a CLI login, or inference bundled inside a coding tool or builder. Speech-to-text, voice synthesis, code, vision, and reasoning. The test is whether you can ship an artifact you keep — code, an app, an audio file, a site, a design, an agent, a skill. Chat-only free tiers are excluded on purpose.

verified 2026-11-12 · about / contribute

Which free LLM API should I use?

Free LLM inference you can build with: 49 providers, 347 models, no card for most entries. Pick by how you get in:

Widest choice: NVIDIA NIM (build.nvidia.com) — 40 free models in one place
Direct API key: Speka — 27 free models, callable with a key
Coding tool: v0 (Vercel) — 6 free models, bundled access

Every provider is listed below, with a detail page for limits and how to get in.

Provider	Free tier	Free models	RPM	TPM	RPD	Day tokens	Context	Timeout	Notes
OpenCode Zen	Promo free (limited time)	Big Pickle, Iceman Free, Inkling 2 Small Free, Crouton Alpha Free, GLM-5.3-Flash Free, DeepSeek V4.1 Flash Free, Beam Free, Flint 40B Free, MiMo-V2.6-Flash Free, Nemotron 3 Ultra Free, Nemotron 3.5 Lightning Free, Nemotron 3.5 Super VL Free, Jev 1.13 Free, MiniMax M3.1 Flash Free, LongCat 2.5 Free	—	—	—	—	—	Not published	[The Grant] OpenCode Zen provides a collection of zero-cost promotional and experimental models. [The Output] Users gain access to free tier inference across multiple models including text and vision variants. [The Process] Models are evaluated and identified via catalog audits and pricing data verification. [The Condition] Availability is subject to limited-time promotional windows, zero-data-retention policies, and user feedback collection requirements.
OpenRouter	Rate-limited free	cohere/north-mini-code:free, dots-studio/dots-3-note-preview:free, google/gemma-4-26b-a4b-it:free, google/gemma-4-31b-it:free, liquid/lfm-2.5-2.6b:free, liquid/lfm-3-1.7b:free, nvidia/nemotron-3-nano-omni-30b-a3b-reasoning:free, nvidia/nemotron-3-ultra-550b-a55b:free, nvidia/nemotron-3.5-lightning:free, nvidia/nemotron-3.5-super-vl:free, recursive/flint-v1.0:free, xingchen-agi/xing4.0-29b-a4b:free, meta/muse-spark-1.3-contributor:free, Crouton Alpha, Lyria 3 Clip Preview, Lyria 3 Pro Preview, Free Models Router, Google: Lyria 3 Pro Preview, Google: Lyria 3 Clip Preview, Inkling 2 Small (free), Thinking Machines: Inkling 2 Small (free), Baidu: ERNIE-5.5-Flash-Preview (free), ERNIE-5.5-Flash-Preview (free)	—	—	—	—	—	Not published	OpenRouter offers 13 permanently free models (:free suffix) with 64K–500K context, translating to 50 requests/day (1,000 requests/day with + purchased credits). No credit card required. Upstream model availability determines live rate limits.
Google AI Studio (Gemini API)	Rate-limited free	gemini-3.5-flash-lite, gemini-3.1-flash-lite, gemma-4-26b-a4b-it, gemma-4-31b-it, gemini-3.5-flash, gemini-3-flash-preview, gemini-3.1-flash-lite-preview, gemini-3.6-flash, gemini-3.7-flash, gemini-3.8-flash, gemini-flash-latest, gemini-flash-lite-latest, gemini-2.5-flash, gemini-2.5-flash-lite, gemini-omni-1.1-flash, gemini-omni-flash-preview	—	—	—	—	—	Not published	Google AI Studio offers 16 free models with up to 1M context. Flagship light models (Gemini 3.5 Flash-Lite, 3.1 Flash-Lite) offer 500 RPD, 15 RPM, and 250K TPM, while open-weight Gemma 4 models (26B, 31B) provide 14.4K RPD, 30 RPM, and 16K TPM. No credit card required. Free-tier prompts and outputs are retained for model training.
Groq	Rate-limited free	llama-3.3-70b-versatile, llama-3.1-8b-instant, meta-llama/llama-4-scout-17b-16e-instruct, openai/gpt-oss-120b, openai/gpt-oss-20b, qwen/qwen3.6-27b, qwen/qwen3.8-27b, meta-llama/llama-prompt-guard-2-22m, compound-beta, whisper-large-v3-turbo, whisper-large-v3, canopylabs/orpheus-v1-english, canopylabs/orpheus-arabic-saudi, ALLaM-2-7b, meta-llama/llama-prompt-guard-2-86m	—	—	—	—	—	Not published	Groq offers 13 free models (9 LLMs, 2 Whisper STT, 2 Orpheus TTS), translating to up to 14,400 text requests/day and 2,000 audio transcriptions/day (~8 hours/day). No credit card required. Limits are pooled at organization level.
Cerebras	Trial credits	gpt-oss-120b, qwen-3.8-27b	—	—	—	—	—	Not published	Cerebras offers $5 trial credit across 2 models (gpt-oss-120b, qwen-3.8-27b with 8K context cap), translating to 1M tokens/day at 2,000+ tokens/sec. Credit card required. Trial credits expire in 30 days.
Cloudflare Workers AI	Rate-limited free	~80 free models (Llama 3.x/4, Qwen, Gemma, DeepSeek-R1 distills, FLUX, Whisper, BGE embeddings), @cf/meta/llama-3.3-70b-instruct-fp8-fast, @cf/deepseek-ai/deepseek-r1-distill-qwen-32b, @cf/qwen/qwen2.5-72b-instruct, @cf/meta/llama-3.1-8b-instruct, @cf/openai/whisper	—	—	—	—	—	Not published	Cloudflare Workers AI is Generally Available with ~80 free models (including Llama 3.3 70B, DeepSeek-R1 distills, Whisper, and FLUX), translating to 10,000 neurons/day (~hundreds of queries or audio minutes). No credit card required. Hard stop at 10,000 neurons until 00:00 UTC daily reset.
Mistral La Plateforme	Rate-limited free	mistral-small-latest, codestral-latest, open-mistral-nemo	—	—	—	—	—	Not published	Mistral provides a free tier for developers to evaluate models. The output includes access to select baseline models via API. The process involves signing up for La Plateforme and utilizing default rate limits. The condition requires compliance with terms of service and staying within specified usage thresholds.
SambaNova Cloud	Trial credits	Meta-Llama-3.3-70B-Instruct	—	—	—	—	—	Not published	1. [The Grant] SambaNova Cloud provides a Free tier offering initial exploratory pay-as-you-go credits. 2. [The Output] Users gain access to select production models under specified free tier parameters. 3. [The Process] The evaluation was conducted by auditing official platform documentation and analyzing API response rate limits. 4. [The Condition] Specific models such as MiniMax-M3.1-Flash are excluded from the free tier offerings.
NVIDIA NIM (build.nvidia.com)	Trial credits	GLM-5, Kimi-2.5, NIM-packaged open models, MiniMax-M3.1-Flash, Whisper Large v3, GPT OSS 20B, Qwen Image, Qwen Image Edit, Llama Guard 4 12B, Llama-3.2-90B-Vision-Instruct, Muse Glimmer 30B, Llama 3.2 11b Vision Instruct, Gemma-4-31B-IT, DiffusionGemma 26B A4B IT, paligemma, DeepSeek V4.1 Flash, FLUX.1-dev, FLUX.2 Klein 4B, FLUX.1-Kontext-dev, FLUX.1-schnell, Laguna XS 2.1, magpie-tts-zeroshot, cosmos-transfer2.5-2b, llama-nemotron-embed-vl-1b-v2, Nemotron 3.5 Lightning 30B A3B, Active Speaker Detection, llama-nemotron-rerank-vl-1b-v2, Nemotron 3 Ultra 550B A55B, Nemotron 3 Nano Omni, llama-3.1-nemotron-safety-guard-8b-v3, nemotron-voicechat, streampetr, bevformer, Nemotron 3.5 Content Safety, sparsedrive, Nemotron 3.5 Super VL 120B A12B, synthetic-video-detector, studiovoice, GLM-5.3-Flash, Step 5 Preview	—	—	—	—	—	Not published	NVIDIA NIM offers 1,000 free trial credits across catalog models (including Nemotron 3.5 Lightning with 1M context), translating to ~40 hours of inference testing. No credit card required; corporate email required for extended credits. Credits tie to a 90-day evaluation window.
Z.AI (Zhipu)	Rate-limited free	GLM-5.1, GLM-4.5-Flash, GLM-4.7-Flash, GLM-4.6V-Flash (vision), GLM-4.6V-Flash	—	—	—	—	—	Not published	Z.AI offers free API access to GLM-5.1, GLM-4.7-Flash, and GLM-4.6V-Flash with 128K context, translating to 15 requests/minute. No credit card required. Shared free tier throttles during peak platform traffic hours.
DeepInfra	Trial credits	deepseek-ai/DeepSeek-V3.1, Qwen/Qwen3-32B, Qwen/Qwen2.5-72B-Instruct, zai-org/GLM-4.7	—	—	—	—	—	Not published	DeepInfra offers $5 signup trial credit across 185+ models (including DeepSeek-V3.1, Qwen 3 32B, and GLM-4.7), translating to ~50M tokens at 200 concurrent requests. No credit card required. Credits remain active until depleted.
GitHub Copilot (Free)	Rate-limited free	Auto-selected (GPT-5 mini, Claude Haiku/Sonnet, Gemini Flash line)	—	—	—	—	—	Not published	GitHub provides Copilot Free with limited features for individual developers to try AI-powered coding assistance. The system delivers restricted agent access and limited AI credit allowances alongside Copilot Chat on GitHub.com and mobile. The implementation uses automated web data extraction, HTML table parsing, and policy analysis scripts via bash. Users receive free access subject to specific AI model availability constraints and organization or enterprise policy governance.
Kiro (AWS)	Rate-limited free	Qwen3 Coder Next, DeepSeek 3.2, MiniMax M2.1, Claude Sonnet 4.5	—	—	—	—	—	Not published	Kiro offers 50 free credits/month for spec-driven agent workflows. Social account or AWS Builder ID required, no credit card required. Unused credits do not roll over to the next month.
Trae (ByteDance)	Trial credits	Frontier models (bundled, auto-selected)	—	—	—	—	—	Not published	Trae offers free inference. The free allowance includes 5,000 monthly autocompletions and a one-time $3 usage allowance. Usage is renewed monthly, and the credits are valid for 90 days. The free tier requires no credit card.
Bolt.new	Rate-limited free	Frontier models (bundled, auto-selected)	—	—	—	—	—	Not published	Bolt offers 1,000,000 tokens per month with a 300,000 daily token cap, translating to bundled frontier model inference for full-stack web and code generation via the Standard agent alongside hosting for up to 333k requests. No credit card required to sign up. Daily tokens reset every 24 hours without rollover, monthly quotas reset on the 1st of each calendar month, and free plan tokens do not roll over.
Lovable	Rate-limited free	Fable 5.1 (Frontier model)	—	—	—	—	—	Not published	1. The Grant: Lovable provides a rate-limited free tier for users building and iterating on applications. 2. The Output: Users get access to bundled frontier models including Fable 5.1 without direct financial cost. 3. The Process: Usage is governed by credits and rate limits determined by the platform's current policy. 4. The Condition: Availability and rate limits are subject to change as outlined in the official documentation.
v0 (Vercel)	Rate-limited free	Frontier models (bundled, auto-selected), Beam 1.0, Beam 1.0 (Free), Glyph Cluster, Nemotron 3.5 Lightning Free, Ampersand v0.1	—	—	—	—	—	Not published	v0 offers a free plan granting 7 messages per day with bundled frontier models, plus a one-time $5 credit bonus for new accounts. Translates to 7 full-stack code and UI generations daily (~210 requests per month). No credit card required to sign up and build. Quotas reset daily at 00:00 UTC with no rollover, and custom model selection is restricted to paid plans.
Kaggle Notebooks	Rate-limited free	Open-weight models you load (GPU runtime)	—	—	—	—	—	Not published	Kaggle offers 30 hours of free GPU accelerator quota (NVIDIA T4 x2 / P100) and 20 hours of TPU v3-8 compute per week, translating to running self-hosted open-weight model inference and fine-tuning across sessions lasting up to 12 hours. Phone verification is required to unlock accelerators and internet access; no credit card required. Compute quota resets weekly, and interactive sessions terminate after 20 to 60 minutes of idle time.
Hugging Face Spaces	Rate-limited free	Any Hugging Face model you load (ZeroGPU), GLM-4.7-Flash	—	—	—	—	—	Not published	Hugging Face Spaces provides 5 minutes of daily ZeroGPU compute on dynamic NVIDIA RTX Pro 6000 Blackwell GPUs (48 GB to 96 GB VRAM) along with free hosting for up to 2 Gradio Spaces. This allows users to deploy and run inference on any compatible Hugging Face model across text, vision, audio, and multimodal tasks with execution limits of 60 to 120 seconds per function call. Access requires creating and logging into a free Hugging Face account with a verified email address, with no credit card required. The 5-minute GPU quota resets 24 hours after the first GPU usage, spaces auto-sleep when inactive, and hosting personal ZeroGPU demos requires an account older than 30 days in good standing.
Cursor (Hobby)	Rate-limited free	Cursor Free Models (Agent/Chat)	—	—	—	—	—	Not published	1. Cursor grants users access to a complimentary Hobby tier for basic AI coding features. 2. The output provides limited agent requests and limited chat requests under rate-limited conditions. 3. The process involves checking the Cursor pricing and models documentation to cross-reference available free offerings. 4. This access remains conditioned upon staying within the specified rate limits of the free tier.
Windsurf	Rate-limited free	Standard Models	—	—	—	—	—	Not published	Windsurf provides a free tier. Users receive a set amount of usage and standard features at no cost. The free tier operates on a rate-limited basis. Usage allowances reset periodically under standard usage conditions.
Devin (Free)	Rate-limited free	SWE-2 (Free), Limited models (auto-selected)	—	—	—	—	—	Not published	Devin offers a Free plan with a light quota to code with agents, free use of SWE-2, unlimited inline edits, and unlimited Tab completions. Account login required, no credit card required. Runs bundled inside the Devin interface on SWE-2 and limited models.
Qoder	Rate-limited free	Frontier models + BYO key	—	—	—	—	—	Not published	Qoder offers free basic code autocompletions and limited Next-Edit-Suggestion (NES) runs. Account login required, no credit card required. Unlimited usage requires bringing your own API keys.
Antigravity CLI (Google)	Rate-limited free	Gemini 3.8 Flash, Gemini 3.7 Flash, Gemini 3.6 Flash, Gemini 3.1 Pro	—	—	—	—	—	Not published	The free tier provides access to select base models for general inference. The output consists of standard text, code, and basic multimodal completions. The process requires standard user authentication via the Antigravity CLI without requiring payment instrument setup. The condition is subject to dynamic rate limits and potential feature restrictions across modalities.
Google Colab	Rate-limited free	Gemini (AI features) + GPU for your code, Lyria 3 Clip Preview, Lyria 3 Pro Preview	—	—	—	—	—	Not published	Google Colab offers free access to hosted Jupyter notebook compute resources including shared GPUs/TPUs and integrated Gemini AI coding features, translating to interactive execution sessions of up to 12 hours with standard system memory. Google account login required (age 18+ for AI features), no credit card required. Resources and rate limits fluctuate dynamically, idle runtimes disconnect, and non-interactive background execution is disallowed.
Figma (Starter)	Rate-limited free	Figma AI, Make, agent	—	—	—	—	—	Not published	Figma Starter offers 150 free AI generation credits/day (up to 500/month). Figma account required, no credit card required. Inference runs bundled inside Figma design files for layer generation and text editing.
Canva (Free)	Rate-limited free	Canva AI, Magic Media	—	—	—	—	—	Not published	Free users are granted an allowance of 20 uses of standard or premium AI tools. The system outputs access to bundled models like Canva AI and Magic Media. The process enforces monthly limits on tools such as Magic Write, consuming a shared allowance. The condition strictly restricts free users from accessing ultra tools like Canva AI 2.0.
Framer (Free)	Trial credits	Framer Agents	—	—	—	—	—	Not published	Framer offers 500 one-time AI trial credits on its Free plan for workspaces without active subscriptions to power Framer Agents, translating to approximately 1 to 3 page generation or design sessions. No credit card is required to sign up and use the free plan. Credits are a one-time trial allotment that does not reset monthly, and agent features pause once the credits are exhausted unless upgraded to a paid plan.
Replit (Starter)	Rate-limited free	Frontier models (bundled, auto-selected)	—	—	—	—	—	Not published	Starter tier grants access to basic development tools. The output includes standard cloud developer workspace capabilities. The process involves creating a free Replit account to immediately provision basic resources. The condition requires operating within standard resource limits and usage policies.
Kilo Gateway	Rate-limited free	cohere/north-mini-code:free, poolside/laguna-s-2.1:free, poolside/laguna-xs-2.1:free, liquid/lfm-2.5-2.6b:free, dots-studio/dots-3-note-preview:free, nvidia/nemotron-3.5-lightning:free, nvidia/nemotron-3.5-super-vl:free, Reflection: Beam 1.0 (free), Crouton Alpha, Stealth: Glyph Cluster (free), Auto Free, Lyria 3 Clip Preview, Lyria 3 Pro Preview, LiquidAI: LFM3-1.7B (free), NVIDIA: Nemotron 3 Nano Omni (free), NVIDIA: Nemotron 3 Ultra (free), Body Builder (beta), OpenRouter Free Models Router, Pareto Code Router, Auto Router, Thinking Machines: Inkling 2 Small (free), Baidu: ERNIE-5.5-Flash-Preview (free)	—	—	—	—	—	Not published	Kilo Gateway offers 7 free models with 64K–256K context, translating to 200 requests/hour per IP address. No credit card required. Model availability reflects upstream provider quotas.
Nous Portal	Rate-limited free	poolside/laguna-s-2.1:free, poolside/laguna-xs-2.1:free, stepfun/step-3.7-flash:free, upstage/solar-mini4:free, reflection/beam-1.0-501b-a23b:free, ornith/ornith-1.5-122b:free	—	—	—	—	—	Not published	Nous Portal offers 5 free Hermes and open-weight models, translating to 50 requests/minute and 500,000 tokens/minute. Free account registration required, no credit card required. Preview models may be adjusted during experimental rounds.
Cohere	Rate-limited free	Command A+, Command A Reasoning, Command R+, Command R, Command R7B, North Small Translate, North Mini Code	—	—	—	—	—	Not published	Cohere provides trial API keys [The Grant] allowing access to chat and embedding models [The Output] by registering an account and utilizing trial keys [The Process] subject to trial rate limits such as 20 requests per minute and 1,000 API calls per month [The Condition].
Requesty	Rate-limited free	nvidia/nemotron-3-super-120b-a12b, google/gemma-4-31b-it, poolside/laguna-xs.2, Gemma 4 31B IT, ling-3.0-tiny, Muse Glimmer 30B, Nemotron 3.5 Lightning 30B A3B, leanstral-1-5, Nemotron 3 Ultra 550B A55B, Nemotron 3 Nano Omni 30B A3B Reasoning, leanstral-1-5@eu, Nemotron 3 Super 120B A12B, Nemotron 3.5 Content Safety	—	—	—	—	—	Not published	Requesty offers 3 free models with 128K context, translating to 200 requests/day per account. Email registration required, no credit card required. Quota resets daily at 00:00 UTC.
OVHcloud AI Endpoints	Rate-limited free	10 free models (anonymous keyless), Qwen3Guard-Gen-0.6B, Qwen3Guard-Gen-8B	—	—	—	—	—	Not published	OVHcloud offers 10 free open-source models, translating to 2 requests/minute per IP keyless (or 400 requests/minute with a registered account). No credit card required for anonymous keyless access. Quota is enforced per client IP.
SiliconFlow	Trial credits	Speech models (unspecified)	—	—	—	—	—	Not published	The Grant: $200 in credits is provided. The Output: Credits can be used for speech models. The Process: Access is granted without credit card requirement. The Condition: Credits are valid for 90 days.
DeepSeek Platform	Trial credits	deepseek-chat, deepseek-reasoner	—	—	—	—	—	Not published	The Grant: DeepSeek offers a granted balance upon account creation. The Output: Users can access API services without immediate paid top-ups. The Process: The platform automatically prioritizes using the granted balance first when both balances are available. The Condition: This applies until the granted balance is exhausted or expires, after which accounts rely on the topped-up balance.
Braintrust	Trial credits	deepseek-v4.1-flash, deepseek-v4-flash-0731, glm-5.3-flash, glm-5.2, kimi-k3, jev, gpt-6-luna, gpt-6-sol, gpt-6.1-sol, gpt-6-astra, gpt-5.6-luna, gpt-5.6-terra, gpt-5.6-sol	—	—	—	—	—	Not published	Braintrust offers $10/month recurring model credit on Starter across 13 built-in models (including GPT-6 series, DeepSeek V4.1, GLM-5.3, and Jev). Work email required, no credit card required. Model credit resets monthly with no rollover.
Deepgram	Trial credits	nova-3, nova-2, flux, aura	—	—	—	—	—	Not published	[The Grant] Deepgram provides $200 in free trial credits upon sign-up. [The Output] Users gain access to public speech-to-text and text-to-speech models including Nova-3, Nova-2, Flux, and Aura. [The Process] Users register for a Pay-As-You-Go account to receive the API key and access the platform. [The Condition] The credit allows for approximately 700 hours of Nova-3 transcription, subject to standard concurrency limits for pre-recorded and streaming requests.
AssemblyAI	Trial credits	Unknown	—	—	—	—	—	Not published	New accounts receive $50 in credits, valid for 90 days, usable across four speech models, and requiring no credit card.
Dots Studio	Trial credits	dots-3-note-preview	—	—	—	—	—	Not published	- The Grant: Users receive a trial credit system accessible via an API key. - The Output: A single available model named "dots-3-note-preview" is provided in the catalog. - The Process: Investigation of JavaScript bundles, API parameters, and documentation revealed context limits and model naming structures. - The Condition: Total input and output token sum cannot exceed the 524,288 token context window limit.
Liquid AI	Trial credits	Liquid AI Free Model	—	—	—	—	—	Not published	1. Liquid AI offers free promotional API credits to new developers upon console signup. 2. The output provides access to evaluate their models via the developer console and API. 3. The process requires creating an account on console.liquid.ai to claim the promotional allocation. 4. The condition is subject to promotional limits and potential expiration once the initial credit balance is exhausted.
Speka	Permanent $0	deepseek-ai/deepseek-v4-flash, deepseek-ai/deepseek-v4-pro, meta/llama-3.1-8b-instruct, meta/llama-3.2-3b-instruct, meta/llama-3.2-11b-vision-instruct, meta/llama-3.2-90b-vision-instruct, meta/llama-3.3-70b-instruct, mistralai/mistral-nemotron, mistralai/mistral-medium-3.5-128b, mistralai/mistral-large-4-1t-a52b, nvidia/nemotron-3-nano-30b-a3b, nvidia/nemotron-3-super-120b-a12b, nvidia/nemotron-3-ultra-550b-a55b, nvidia/nemotron-3.5-bolt-e4b, nvidia/nemotron-3.5-lightning-30b-a3b, nvidia/nemotron-3-embed-1b, nvidia/nemotron-nano-9b-v2, nvidia/nemotron-nano-12b-v2-vl, nvidia/llama-3.3-nemotron-super-49b-v1.5, openai/gpt-oss-20b, openai/gpt-oss-120b, poolside/laguna-xs-2.1, stepfun-ai/step-3.7-flash, stepfun-ai/step-5-preview, minimaxai/minimax-m3, z-ai/glm-5.2, black-forest-labs/flux.1-dev	—	—	—	—	—	Not published	[The Grant] Speka provides a recurring $1.00 monthly model usage allowance on its Free plan ($0/mo) that refreshes every month. [The Output] Developers can generate completions, embeddings, and images across all open catalog models with 1 API key and a rate limit of 10 requests per minute (RPM). [The Process] Users sign up with email and password via speka.me/signup without requiring a credit card or upfront payment, and instantly create an API key in the dashboard. [The Condition] Requests are billed against the $1 monthly credit at standard per-token model rates; once the monthly allowance is exhausted, requests return HTTP 402 until the next monthly reset or paid upgrade.
Alibaba DashScope (Bailian)	Trial / Free tokens	Qwen3.7 Plus, DeepSeek V4.1 Flash 1017, GLM-5.3, Qwen3.6 Plus, Wan2.7 Image, HappyHorse 1.1 Reference-to-Video, Qwen3.6 Flash, Qwen3.7 Max, DeepSeek V4 Pro, MiniMax-M2.5, Kimi K2.6, Kimi K2.5, Kimi K2.7 Code, HappyHorse 1.1 Text-to-Video, Qwen Image 2.0, Wan2.7 Image Pro, HappyHorse 1.1 Image-to-Video, DeepSeek V4 Pro 0813, Qwen3.8 Flash, Qwen3.8 Max, Qwen Image 2.0 Pro	—	—	—	—	—	300s	The Grant: Alibaba Cloud Model Studio (Bailian/DashScope) provides free token allotments upon account activation across its Qwen and DeepSeek model fleet. The Output: Access to high-throughput OpenAI-compatible endpoints with up to 1M token context windows. The Process: Register an Alibaba Cloud account and generate an API key in the Model Studio console. The Condition: Free token allocations apply for designated promotional windows with phone verification.
ModelScope	Rate-limited free	Qwen3 30B A3B Thinking 2507, Qwen3 235B A22B Instruct 2507, Qwen3 30B A3B Instruct 2507, Qwen3-235B-A22B-Thinking-2507, Qwen3 Coder 30B A3B Instruct, GLM-4.5, GLM-4.6	—	—	—	—	—	180s	The Grant: ModelScope (Alibaba's open-source model community) offers free hosted inference for community and Qwen open-weights models. The Output: Standard OpenAI-compatible completion endpoints with 262K token context windows. The Process: Create an account on ModelScope and retrieve an API access token. The Condition: Rate-limited to community usage tiers with fair-share queueing.
AMD Developer Cloud	Free Developer Tier	MiniCPM5-2B, Qwen3.8 Flash Next, DeepSeek V4 Flash 0731, MiniCPM-V 4.7 1B, MiniCPM-V 4.7 35B-A3B, Qwen3.8 27B, MiMo-V2.6-Flash, DeepSeek V4.1 Flash, Gemma 4 31B IT, GLM-5.3-Flash	—	—	—	—	—	120s	The Grant: AMD Developer Cloud provides free inference access to open frontier models running on AMD Instinct ROCm GPU clusters. The Output: OpenAI-compatible chat endpoints supporting DeepSeek, Qwen, and MiniCPM variants. The Process: Register an AMD Developer account to access the inference API console. The Condition: Subject to ROCm cluster capacity and developer fair-use quotas.
InferX	Free Developer Tier	Qwen3.6 35B A3B FP8, deepseek-v4-flash, Ornith-1.0-35B-FP8, Qwen3.6 27B FP8, mimo-v25, Qwen3 Coder Next FP8, Agents-A1, Devstral-2-123B-Instruct-2512-int4-AutoRound, Qwen3-Embedding-8B, Qwen3.6-35B-A3B-fp8-no-thinking, Gemma 4 31B IT FP8, Qwen3-Coder-Next-FP8-no-thinking	—	—	—	—	—	120s	The Grant: InferX provides zero-cost inference endpoints for cutting-edge open models including Qwen 3.6, Gemma 4, and DeepSeek variants. The Output: Ultra-low latency OpenAI-compatible completions with streaming and structured output support. The Process: Sign in to generate an API key and query endpoint URLs directly. The Condition: Rate-limited for community development and prototyping.
Kimi For Coding (Moonshot AI)	Free Developer Tier	Kimi K3.1 Pro Highspeed, Kimi K3-256K, Kimi K3.1 Pro, Kimi K3, kimi-for-coding, Kimi For Coding HighSpeed	—	—	—	—	—	180s	The Grant: Moonshot AI provides free developer inference for Kimi K3 and coding models with context windows up to 1M tokens. The Output: OpenAI-compatible coding endpoints supporting deep reasoning and agentic workflows. The Process: Authenticate via Kimi coding portal to retrieve API credentials. The Condition: Community rate limits and high-concurrency throttling during peak periods.
SCNet	Free Developer Tier	Kimi K2.7 Code, Qwen3.8 Flash, MiniMax-M2.5, DeepSeek V4 Flash, DeepSeek V4 Pro 0813, GLM-5.3, MiniMax-M3, MiniMax-M2.7, Qwen3.8 Max, DeepSeek V4 Pro, Kimi K2.6, Kimi K2.5	—	—	—	—	—	180s	The Grant: SCNet provides national high-performance computing open inference tokens for domestic LLMs including Qwen 3.8 and DeepSeek V4 Flash. The Output: OpenAI-compatible chat endpoints with up to 1M token context windows. The Process: Register on SCNet portal and generate developer API keys. The Condition: Valid identity verification and developer quota rules apply.
AIHubMix	Promo free	GLM-4.7-Flash (free), Nemotron 3 Ultra 550B A55B (free), Nemotron 3.5 Content Safety (free), Qwen3.8 Max (free), GLM-5.2 (Coding, free), Ling 3.0 Flash (free), Coding MiniMax M2.7 (Free), Nemotron 3.5 Lightning 30B A3B (free), GLM-5.3 (Coding, free), Qwen3.8 Flash (free), Nemotron 3 Nano Omni 30B A3B Reasoning (free), Coding GLM 5.1 (free), GLM-5.3-Flash (Coding, free), Nemotron 3 Super 120B A12B (free)	—	—	—	—	—	120s	The Grant: AIHubMix provides a suite of zero-cost promotional models including GLM-4.7 Flash, Nemotron 3 Ultra, and Qwen 3.8 Max. The Output: OpenAI-compatible completion endpoints with long context support up to 1M tokens. The Process: Sign up for an AIHubMix account to retrieve free tier API credentials. The Condition: Rate-limited free promotional usage subject to provider token availability.