August 2026 Edition, Verified via OpenRouter Live Data

The Complete Guide to Free LLM APIs
and Low-Cost Inference in 2026

Token costs are climbing. Your Hermes Agent, OpenClaw, or Open Code project should not stall because of a billing alert. Here is every provider offering free API access right now, what the real limits are, and the cheapest unlimited-style subscriptions when you need to go further.

By Pierre Bradshaw August 20, 2026 Updated Every Month

Why this matters: A multi-step agent workflow can run $40 to $200 a month at list price. Stacking free tiers intelligently covers a large share of that for $0. The remaining calls, where you actually need frontier-level reasoning, is where you spend. This guide shows you both sides, refreshed with live OpenRouter data as of August 20, 2026.

Quick Glossary: RPM Means Requests Per Minute, RPD Means Requests Per Day

Terms you will see throughout this guide

RPM = Requests Per Minute
The number of API calls you can send in any 60-second window before the provider throttles or rejects the next one with a 429 error.

RPD = Requests Per Day
The total number of API calls allowed in a rolling or calendar day, regardless of how you space them out.

TPM
Tokens Per Minute. A separate ceiling some providers use instead of, or alongside, RPM, since one request can contain very different amounts of text.

429 Error
The HTTP status code returned when you exceed a rate limit. It means slow down, not you are banned.

Why the units matter: A provider advertising "1,000/day" and one advertising "1,000/month" are not comparable. Throughout this guide, every number is labeled RPD (per day) or RPM (per minute) so you can compare providers on equal footing. Where a provider sets different caps per model, we list the specific caps for each top model rather than a single blended number.

🚀

Lean team, Building Fast

Jump to the Recommended Stack. Google AI Studio, NVIDIA NIM, and OpenRouter cover most MVPs at $0.

🤖

Running Hermes Agents

See the NVIDIA NIM card. It free-hosts GLM-5.2 with no credit ceiling, gated only by a 40 RPM cap.

💸

Scaling Past Free Tiers

Featherless and the GLM Coding Plan give flat-rate unlimited usage starting at $18 to $25/month.

🧑‍💻

Coding in Cursor or VS Code

See what Cursor, Codex, and Antigravity actually give you free before you touch a raw API.

Quick Reference: All Free API Tiers

No credit card, no trial period limits, verified August 20, 2026

Provider	Top Free Models	Daily / Monthly Limit	Rate Limit	Card Required	Specialty Modalities	How Long Free
Google AI Studio	Gemini 3.6 Flash, Gemini 3.5 Flash-Lite, Gemini 3.1 Flash-Lite	Flash-Lite: up to 1,500 RPD; Flash: ~250-500 RPD; Pro models paid-only since Apr 1, 2026	Flash-Lite ~15 RPM; Flash ~10-15 RPM, 250K TPM	No	Vision, Audio, Video, PDF input	Ongoing; Google has cut quotas before without notice
NVIDIA NIM	GLM-5.2, Nemotron 3 Ultra 550B, Nemotron 3.5 Lightning	No credit or token cap; gated only by rate limit	40 RPM shared cap, varies with model load	No	Vision, embeddings, safety/guard models	Permanently free; NVIDIA removed its old credit ceiling
Groq	Llama 3.1 8B Instant, Llama 4 Scout, GPT-OSS 120B	14,400 RPD (8B); 1,000 RPD (70B/Scout/GPT-OSS); 250 RPD (Compound)	30-60 RPM; 6,000-30,000 TPM depending on model	No	Whisper Large v3 & Turbo STT (2,000 audio req/day)	Free Forever tier
Cerebras	gpt-oss-120b, Gemma 4 31B	$5 in credits, expires 30 days after granted	5 RPM, 30K TPM, 1M TPD per model	Yes	Fastest raw inference speed (2,000+ tok/sec)	No longer a true free tier as of Aug 17, 2026; card required to unlock a 30-day credit trial
OpenRouter	GLM-5.2, NVIDIA Nemotron 3 Ultra 550B, Poolside Laguna S 2.1	50 RPD (no balance); 1,000 RPD (with $10+ lifetime credit)	20 RPM	No	Vision and tool calling on select free models	Ongoing; 25+ free models rotate constantly
Mistral La Plateforme	Mistral Small 3.2, Devstral Small, Codestral	1 req/sec (Experiment tier)	~1 RPS	Phone verify	Vision (Pixtral), Code (Codestral)	Ongoing Experiment tier
GitHub Models	GPT-5-mini, Llama 4 Scout, DeepSeek-R1	10-150 RPD (model class dependent)	10-15 RPM	No	Vision on GPT-4.1 / o4-mini class models	Ongoing for GitHub account holders
SambaNova Cloud	Llama 3.3 70B, DeepSeek-V3.1, gpt-oss-120b	Model-specific RPM caps	10-30 RPM	No	Llama Vision (rate limited)	Ongoing free tier
Hugging Face Inference	Llama 3.2 11B Vision, Qwen2.5, FLUX.1 Dev	$0.10/month in credits (free users)	Cold-start dependent	No	Image gen, audio, STT	Ongoing (shared infra)
Cohere	Command A, Command R7B, Rerank 3.5	1,000 API calls/month	20 RPM	No	Embed v4, Rerank, STT (5 RPM)	Ongoing, non-commercial only
DeepSeek	DeepSeek-V4-Flash, DeepSeek-R1	Signup credit varies by promo	Low RPM without paid tier	After credit	1M context, reasoning mode	One-time credit, then pay-as-you-go
Cloudflare Workers AI	Llama 4 Scout, Mistral 7B, Gemma models	10,000 neurons/day	Standard Workers limits	No	Whisper STT, text-to-image, embeddings	Ongoing free plan
Fireworks AI	Apriel 1.6 15B Thinker, DeepCoder 14B, Sarvam M	$1 signup credit, plus 6 permanently free models	10 RPM without card; 6,000 RPM with one	No	Image, speech, embeddings on paid tier	6 models free forever; $1 credit is one-time
Novita AI	Llama 3.2 1B, Qwen2.5-7B, GLM-4-9B	~$0.50 signup voucher (1yr); $10 per referral, up to $500	~60 RPM (third-party reported)	No	Image gen, embeddings (BGE-M3)	5 models at $0/token ongoing; referral credits stack

Important: Cerebras eliminated its no-card free tier on August 17, 2026. A payment method is now required to unlock $5 in credits that expire after 30 days; this is a trial, not a free tier. Together AI also discontinued free signup credits in July 2025 and now requires a $5 minimum prepaid purchase, so it does not appear in this table. Free tier limits and model names change constantly elsewhere too. Verify current model IDs and quotas on each provider's own docs page before wiring anything into production.

Provider Deep Dives

Top models ranked using live OpenRouter usage and Artificial Analysis benchmark data, pulled August 20, 2026

NVIDIA NIM
build.nvidia.com
No Card
NVIDIA hosts genuinely frontier-class open weight models on its own silicon, including Zhipu AI's GLM-5.2, to prove out NIM as the default runtime for open models. NVIDIA removed its old consumable-credit system, so the free tier is now permanent and gated only by rate limit, not by a shrinking credit balance.
Free Usage Cap
None. No credits, no token ceiling, no monthly quota
Rate Limit
40 RPM cap, varies by model load per NVIDIA's own FAQ
GLM-5.2 Context Window
Up to 1,000,000 tokens (model card spec)
Duration
Permanently free; the RPM ceiling is not raisable on the free tier
z-ai/glm-5.2nvidia/nemotron-3-ultra-550bnvidia/nemotron-3.5-lightning
GLM-5.2 tops OpenRouter's weekly usage and coding index (68.8) among all free models on NVIDIA's catalog as of August 2026.
Vision, EmbeddingsData may be used for training

          Access: build.nvidia.com/z-ai/glm-5.2. Sign up for a free permanent API key, no credit card and no credit balance to track. The 40 RPM cap applies across the catalog and NVIDIA's own developer forums confirm it cannot be raised on the free tier. An agent's iterative tool calls can still burn through 40 RPM fast, leading to lag or 429 errors on complex tasks.

          Do not use for confidential or client data. NVIDIA's free-tier terms permit using inputs and outputs to improve their own models.
        

Google AI Studio
ai.google.dev
No Card
Google gives this away to convert developers onto Vertex AI billing. Gemini 3.6 Flash, released July 20, 2026, is now the flagship free-tier model, with Gemini 3.5 Flash still available but effectively superseded. Pro-series models (2.5 Pro, 3.1 Pro) left the free tier entirely on April 1, 2026 and are paid-only now.
Gemini 3.6 Flash
~10-15 RPM, 250K TPM
Flash-Lite RPD
Up to 1,500 requests per day (RPD), highest daily cap of any free model
Context Window
1,048,576 tokens on Flash-class models
Pro Models
Paid-only since April 1, 2026, no free-tier access at all
gemini-3.6-flashgemini-3.5-flash-litegemini-3.1-flash-lite
Correction from prior editions: "Gemini 3 Flash-Lite" is not a real model ID and Gemini 3.1 Pro was never actually free. The current free lineup is Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.1 Flash-Lite; Pro-series models are paid-only.
VisionAudio InputVideo Input
Access: ai.google.dev → Get API Key. Google account only, no card.
Watch out: Google has cut free quotas before without warning; Pro-class models get no free access at all as of April 2026.

Groq

console.groq.com

No Card

Groq built custom LPU chips and needs developers to feel the speed advantage firsthand. Once you experience 300 to 3,000+ tokens/second, slower frontier APIs feel sluggish. Their hardware business is the real product.

Best RPD

14,400 (Llama 3.1 8B Instant)

70B / Scout RPD

1,000/day each

Speed

300-3,000+ tokens/sec

Rate Limit

30-60 RPM, model dependent

meta-llama/llama-4-scoutllama-3.3-70b-versatileopenai/gpt-oss-120b

Whisper Large v3 STT (Free)Ultra-Low Latency

Access: console.groq.com → Sign up → API Keys.
Note: Each model has its own RPD ceiling; the 8B Instant model gets by far the highest daily allowance.

Cerebras

cloud.cerebras.ai

Card Required

Cerebras uses wafer-scale chips and needs workloads that showcase raw inference speed. On August 17, 2026, Cerebras eliminated its no-card free tier entirely, replacing it with a credit-based trial that requires a payment method to unlock.

Speed

2,000+ tokens/sec

Rate Limit

5 RPM, 30K TPM, 1M TPD per model

Trial Credit

$5, expires 30 days after granted

Card Required

Yes, to activate the credit trial at all

gpt-oss-120bgemma-4-31b

Correction from prior editions: Cerebras no longer offers a genuinely free, no-card tier. As of August 17, 2026 it requires a payment method to unlock a time-limited $5 credit trial across only two models.

Fastest Free InferenceLong Context RAG

Access: cloud.cerebras.ai → Sign up. Email only.
Best for: Quick speed tests; move to the $10 tier fast if you need Cerebras for real work.

OpenRouter
openrouter.ai
No Card
OpenRouter monetizes by routing paid calls across providers and taking a small margin. Free models are subsidized by labs who want developer adoption. One API key for 25+ rotating free models is the fastest way to test many models side by side. The roster has shifted noticeably since our last check: GLM-5.2 and NVIDIA's Nemotron family now dominate free usage, with DeepSeek R1 and Qwen3 Coder no longer in the top-weekly free list.
Free Requests, No Balance
50 requests per day (RPD)
Free Requests, $10+ Balance
1,000 requests per day (RPD)
Rate Limit
20 requests per minute (RPM)
API Style
OpenAI-compatible
z-ai/glm-5.2:freenvidia/nemotron-3-ultra-550b:freepoolside/laguna-s-2.1:free
Live-pulled from OpenRouter's list-models endpoint sorted by top-weekly usage and coding index, August 20, 2026.
Vision (select models)Tool Calling (select models)
Access: openrouter.ai → Sign up → API Keys. Use any model ID ending in :free, or see the Free Router section below. The 50 vs. 1,000 RPD split applies to your total free-model usage across ALL free models combined, not per model.

Mistral La Plateforme

console.mistral.ai

Phone Verify

Mistral's Experiment tier is top-of-funnel for a European lab competing with OpenAI and Anthropic. Startups can also apply for grants up to $30,000 in credits through Mistralship.

Rate Limit

~1 req/sec

All API Models

Yes, incl. Codestral

EU Data Residency

Yes

Startup Grant

Up to $30K via Mistralship

mistral-small-3.2devstral-smallcodestral

Vision (Pixtral)EU GDPR Compliant

Access: console.mistral.ai → Sign up → Experiment tier. Phone SMS verification required.

GitHub Models

github.com/marketplace/models

No Card

Microsoft owns GitHub and Azure OpenAI. GitHub Models is the discovery layer that funnels production usage toward Azure.

Total Models

45+

High-Tier RPD

10-50/day

Low-Tier RPD

up to 150/day

Per-Request Limit

Model dependent

openai/gpt-5-minimeta/llama-4-scoutdeepseek/deepseek-r1

Vision (GPT class)OpenAI-Compatible

Access: github.com/marketplace/models → Generate a token with Models:read scope. Base URL: https://models.inference.ai.azure.com

SambaNova Cloud

cloud.sambanova.ai

No Card

SambaNova built RDU chips as an NVIDIA alternative. Rate limits are per-model rather than a single flat cap, so smaller models get far more headroom than 405B-class ones.

Small Models

Up to 30 RPM

70B Models

~20 RPM

405B Models

~10 RPM

Speed

RDU-accelerated

Meta-Llama-3.3-70B-InstructDeepSeek-V3.1gpt-oss-120b

Llama Vision (rate limited)OpenAI-Compatible

Access: cloud.sambanova.ai → Sign up → API Key (starts with sn-).

Hugging Face Inference

huggingface.co

No Card

HuggingFace is the GitHub of AI models. Free inference keeps developers on-platform, driving traffic toward paid Inference Endpoints and PRO plans. Free users get a small monthly credit; PRO users get 20x more.

Free Monthly Credit

$0.10

Model Variety

Thousands

API Format

HF Inference (not OAI)

Cold Starts

Common on rare models

Llama-3.2-11B-Vision-InstructQwen2.5-72B-InstructFLUX.1-dev

Image Gen (FLUX.1 Dev)Audio / STT

Access: huggingface.co → Settings → Access Tokens → Create read token. Not OpenAI-compatible.

Cohere

dashboard.cohere.com

No Card

Cohere is enterprise-focused and uses the free trial key to let developers validate RAG pipeline quality before committing. Rerank and Embed remain best-in-class for retrieval at zero cost. Cohere also shipped North Mini Code, its first agentic coding model, free on OpenRouter.

Monthly Calls

1,000 calls/month

Rate Limit

20 RPM

Context

Up to 256K

Use Case

RAG / Enterprise

command-arerank-3.5embed-v4

Rerank / Embed v4 / STTNon-commercial only

Access: dashboard.cohere.com → API Keys → Trial key. Resets monthly. Not for commercial production use.

What Cursor, Codex, and Antigravity Actually Give You Free

The three coding environments founders ask about most, updated for August 2026

Tool	Free Tier	What's Included	Paid Trigger
Cursor	Hobby plan, $0	Limited Agent requests and limited Tab completions, no credit card. Cursor no longer publishes a fixed numeric quota for Hobby, so treat any specific number you see elsewhere as unreliable	Pro is $20/mo with unlimited Auto mode that never consumes credits; Pro+ $60/mo; Ultra $200/mo
OpenAI Codex	Bundled with ChatGPT Free and Go plans	Codex extension works in VS Code with a ChatGPT login; free accounts get reduced usage caps vs. Plus, Pro, Business, Enterprise	Higher limits and priority require ChatGPT Plus ($20/mo) or above
Google Antigravity	Free Individual tier	Google's own pricing page describes the free tier as having "Basic weekly rate limits" without publishing exact numbers; ships with Gemini 3.5 Flash access	Higher tiers advertise "more generous rate limits," also without published figures
Cline (VS Code extension)	Extension is free forever, open source	No tokens of its own. Bring your own key (BYOK) from any provider in this guide, or use its built-in usage-billing option where a small number of models are tagged FREE	Usage is billed by whichever provider key you connect, not by Cline itself

Practical takeaway: None of these three coding tools publish a stable, verifiable free-tier number anymore, a trend across the industry as usage-based abuse increased. Budget assuming "limited" means "enough to evaluate, not enough to rely on," and pair them with a free API key from this guide via Cline for predictable capacity.

The OpenRouter Free Models Router

One endpoint, zero cost, automatic model selection

OpenRouter ships a special model slug called openrouter/free that removes the need to pick a specific free model yourself.

How It Works

Setup

Set model to openrouter/free in any OpenAI-compatible request. No other code changes needed.

Selection

The router picks from OpenRouter's 25+ available free models per request, filtering automatically for whichever capability your request needs, such as vision, tool calling, or structured outputs.

Context

200,000 tokens on the router endpoint itself, since it must work across every model it might route to.

Rate Limits

Same as any other free model: 50 requests/day by default, up to 1,000/day once you hold a $10+ credit balance, and 20 RPM across all users.

Trade-off

You do not know which model answers each request, so output style and quality vary between calls. Treat it as a discovery and testing tool, not a stable production endpoint.

The Recommended Free Stack

How to cover most of your agent workload at $0/month

Agent Workflow Stack (August 2026)

Default Route

Google AI Studio (Gemini 3.5 Flash) for general chat, document Q&A, and multimodal tasks. Rate-limited but the strongest free multimodal option available.

Hermes / Agent Runtime

NVIDIA NIM for running Hermes Agent, OpenClaw, and Open Code against GLM-5.2, a permanently free tier with no credit balance to track. Watch the 40 RPM ceiling on complex multi-step tool calls, since that is the only real constraint left.

Real-Time/Voice

Groq for latency-sensitive flows: voice agents, streaming chat. 300 to 3,000+ tokens/sec plus free Whisper STT.

Batch Speed

Cerebras now requires a card for even its trial tier after the August 17, 2026 change, so treat it as optional rather than part of your free stack. Fireworks AI is the better free-forever alternative for speed testing, with 6 permanently free models and no card required.

Routing Layer

Cline (free, open source) to manage which provider handles which task inside your IDE, without paying for the tool itself.

RAG Pipeline

Cohere trial key for Rerank 3.5 and Embed v4. Best-in-class retrieval at zero cost, capped at 1,000 calls/month.

Overflow / Frontier

GPT-5.6 Luna pay-as-you-go for the share of calls needing more than free tiers allow; at $0.20/$1.20 per million tokens it posts a 71.4 coding index, nearly matching pricier tiers at a fraction of the cost.

Low-Cost Subscriptions for Unlimited Open Weights

What you actually pay, what it includes, and the top models in each plan, verified August 2026

These are flat-fee subscriptions, not pay-per-token pricing. Use these once you outgrow free tier rate limits and need guaranteed, repeatable throughput for a running agent.

Hugging Face PRO

$9/mo

GLM Coding Lite

$18/mo

$18

Featherless Premium

$25/mo

$25

GLM Coding Pro

$72/mo

$72

GLM Coding Max

$160/mo

$160

Provider / Plan	Price	What's Included	Top 3 Open Weight Models (August 2026)	Best For
Featherless Premium	$25/mo	Unlimited monthly requests, 4 concurrent units, access to any model in the catalogue including Kimi and GLM, private/anonymous usage, no logs	Kimi K2, GLM-5.3, DeepSeek V4 Flash	Founders who want one flat fee for any open model, any size
Featherless Token-Based Business	$25+/mo	Credit-based monthly billing that scales with usage, 8 concurrent units, 256K context, one agent sandbox included	DeepSeek V4 Pro, GLM-5.3, Nemotron 3 Ultra	Running persistent agents (Hermes Agent, OpenClaw) at scale
GLM Coding Lite	$18/mo (~$12.60/mo annual)	~80 prompts per 5-hour cycle, ~400/week, 100 MCP calls/month, flagship GLM access via Claude Code, Cline, or OpenCode	GLM-5.3, GLM-5.2, GLM-4.7	Day-to-day coding on small to mid repos
GLM Coding Pro	$72/mo (~$50.40/mo annual)	~400 prompts per 5 hours, ~2,000/week, 1,000 MCP calls/month, priority queueing	GLM-5.3, GLM-5.2, GLM-4.7	Heavier individual development workloads
GLM Coding Max	$160/mo (~$112/mo annual)	~1,600 prompts per 5 hours, ~8,000/week, 4,000 MCP calls/month, peak-hour priority	GLM-5.3, GLM-5.2, GLM-4.7	Teams or power users running agents continuously
Hugging Face PRO	$9/mo	8x higher ZeroGPU quota, 25 min/day H200 compute, 1TB private storage, higher inference credits	Qwen2.5-72B, Llama 3.3 70B, FLUX.1-dev	Prototyping without cold-start delays

Correction from prior editions: Featherless quietly dropped its old $10 Basic tier and 4-tier ladder up to $200/mo Max. As of August 2026, the entry consumer plan is Featherless Premium at $25/mo with 4 concurrent units, and heavier usage now routes through credit-based Business plans starting at $25+ rather than a fixed $100/$200 Agent Standard/Max ladder. Also new: Z.ai shipped GLM-5.3 on August 14, 2026, now the flagship model across all three GLM Coding Plan tiers, topping Terminal-Bench among open models.

Specialty Models: Image, Video, TTS/STT

Free access beyond text generation

Modality	Provider	Free Model	Free Limit	Access
Image Generation	Hugging Face	FLUX.1 Dev	$0.10/month credit, shared infra	HF Inference API, read token
Image Generation	Cloudflare Workers AI	Stable Diffusion XL	10K neurons/day	Workers AI API
Speech-to-Text	Groq	Whisper Large v3 / Turbo	2,000 audio req/day	console.groq.com audio endpoint
Speech-to-Text	Cloudflare Workers AI	Whisper	Included in 10K neurons/day	Workers AI API
Speech-to-Text	Cohere	Audio Transcriptions	5 RPM (trial key)	Cohere trial key
Text-to-Speech	ElevenLabs	Multilingual v2 / Flash / Turbo	10,000 credits/month (~10 min audio), no commercial license	elevenlabs.io free plan
Vision / Multimodal	Google AI Studio	Gemini 3.5 Flash	~15 RPM, up to 1,500 RPD	ai.google.dev
Vision / Multimodal	OpenRouter	Select :free vision models	20 RPM / 50-1,000 RPD	openrouter.ai
Embeddings	Cohere	Embed v4	Included in 1,000 calls/mo trial	Cohere trial key
Embeddings	Cloudflare Workers AI	BGE / Multilingual embeddings	10K neurons/day	Workers AI API
Embeddings	Novita AI	BGE-M3	$0/token, ongoing	api.novita.ai OpenAI-compatible endpoint
Reranking (RAG)	Cohere	Rerank 3.5	Included in 1,000 calls/mo trial	Cohere trial key
Video Generation	Z.ai / GLM	CogVideoX-3 (paid, low cost)	Not free, roughly $0.20-0.40/video	docs.z.ai

Note: Free video generation remains essentially nonexistent as of August 2026. If your project needs video, budget for this separately.

Benchmarks: Top Free Models Right Now

5 leading open weight models plus the top free Gemini models, live from OpenRouter's Artificial Analysis data, August 20, 2026

Model	Type	Coding Index	Intelligence Index	Agentic Index	Context Window	Best Used For
GLM-5.2 (Z.ai)	Open Weight, Free	68.8	52.6	45.7	256,000 tokens on OpenRouter (1M native)	Long-horizon coding and agent workflows; the top free open weight model by usage and score
NVIDIA Nemotron 3 Ultra 550B	Open Weight, Free	49.3	38.3	27.5	1,000,000 tokens	Long-context orchestration tasks where a 1M window matters more than raw coding score
Poolside Laguna S 2.1	Open Weight, Free	70.2% Terminal-Bench 2.1	Not yet indexed	Purpose-built coding agent	262,144 tokens	Dedicated coding agent tasks; newest top-usage free model on OpenRouter
Google Gemma 4 31B	Open Weight, Free	43.4	29.7	14.4	262,144 tokens	General reasoning with vision input at a smaller compute footprint
NVIDIA Nemotron 3 Super 120B	Open Weight, Free	37.7	25.7	8.8	262,144 tokens	Complex multi-agent applications needing compute efficiency over raw benchmark scores
Gemini 3.5 Flash (Google)	Closed, Free Tier	70.1	52.0	39.7	1,048,576 tokens	The best all-around free model for coding agents, multimodal input, and everyday reasoning
Gemini 3.5 Flash-Lite (Google)	Closed, Free Tier	49.3	37.4	27.2	1,048,576 tokens	High-volume, latency-sensitive subagent tasks inside larger multi-agent workflows

What changed since last month: The free model roster on OpenRouter shifted meaningfully. GLM-5.2 and NVIDIA's Nemotron 3 family now dominate free usage and benchmark scores, while DeepSeek R1, Qwen3 Coder 480B, and Kimi K2 have rotated off the current top-usage list of free models. Gemini 3.5 Flash remains the strongest fully free closed model, essentially tying GLM-5.2 on coding while posting a stronger overall intelligence score.

Fallback Benchmarks: Low-Cost Frontier Models

Where to go when free limits run out, live pricing and benchmarks from OpenRouter, August 20, 2026

Model	Price (In/Out per 1M)	Coding Index	Intelligence Index	Agentic Index	Context Window	Best Used For
Gemini 3.5 Flash	$1.50 / $9.00	70.1	52.0	39.7	1,048,576 tokens	Best cost-to-capability ratio for coding agents and multimodal work
Gemini 3.6 Flash	$1.50 / $7.50 (standard); $0.75 / $3.75 batch	Not separately indexed on current catalog	Strong for its price tier	Google's current recommended default, succeeding 3.5 Flash	1,048,576 tokens	Everyday business tasks, drafts, summarization at low frontier-adjacent cost
GPT-5.6 Luna	$0.20 / $1.20	71.4	52.3	46.9	1,050,000 tokens	Best value in this table; near-Terra coding quality at roughly a tenth of the price
Claude Sonnet 5	$2.00 / $10.00	71.5	55.3	49.7	1,000,000 tokens	Highest coding accuracy per dollar among named frontier chat models; strong default for serious agent coding
Grok 4.5	$2.00 / $6.00 (under 200K context; 2x above)	72.4	55.8	48.9	500,000 tokens	Token-efficient coding agent with the highest coding index of this group at its price point
GPT-5.6 Terra	$2.00 / $12.00	76.7	56.6	50.2	1,050,000 tokens	Highest coding and agentic scores of any mid-priced model in this comparison

How to read this, updated: Pricing corrections this month matter. GPT-5.6 Terra is actually $2/$12, not the $2.50/$15 figure we published previously, and GPT-5.6 Luna is dramatically cheaper than earlier reported at $0.20/$1.20 while still posting a 71.4 coding index, nearly matching Claude Sonnet 5's 71.5 at a fraction of the cost. Terra currently posts the highest coding (76.7) and agentic (50.2) scores of this group. Grok 4.5 remains the value pick for token-efficient coding agents, and Sonnet 5's $2/$10 pricing has held steady since June.

5 Rules for Staying Inside Free Limits

1. Cache repeated answers. Most rate limits count every call, not unique queries. A cached response is worth hundreds of API calls saved.

2. Right-size your model. Classification and summarization do not need a 70B model. Save the largest free models for reasoning and generation steps.

3. Trim your context aggressively. Since most limits measure tokens, a lean 400-token system prompt beats a 2,000-token one every time.

4. Route by task type. Interactive chat to Groq, bulk processing to Fireworks AI's free models, multimodal to Google AI Studio, Hermes-style agents to NVIDIA NIM. Stacked limits multiply your effective free quota.

5. Watch RPM before RPD. Most 429 errors come from bursting requests too fast, not from hitting a daily ceiling. Add small delays between agent tool calls to stay under the per-minute limit.

Disclosure: Free tier limits and model rosters change without warning. Cerebras eliminated its no-card free tier entirely on August 17, 2026 and now requires a payment method to unlock a time-limited $5 credit trial. Together AI discontinued free signup credits in July 2025 and now requires a $5 minimum prepaid purchase. Google has cut Gemini free quotas before without notice. NVIDIA NIM's free tier is permanently free with no credit ceiling but its terms still permit using free-tier inputs and outputs to improve their models. OpenRouter's higher daily request cap (1,000 RPD vs. 50 RPD) depends on maintaining a $10+ funded balance, not usage history. Cursor, Codex, and Antigravity no longer publish fixed free-tier quotas. Always confirm current RPM, RPD, and per-model caps on the provider's own docs before routing business-critical or client data through a free API. This guide is republished on the 5th of each month.