Skip to content

Instantly share code, notes, and snippets.

@krishnakanthb13

krishnakanthb13/AI Providers.md

Last active March 13, 2026 10:22
Show Gist options

  • Select an option

    Learn more about clone URLs
  • Save krishnakanthb13/9ae949958a0bd021d06bf415d16012ab to your computer and use it in GitHub Desktop.

Select an option

Learn more about clone URLs
Save krishnakanthb13/9ae949958a0bd021d06bf415d16012ab to your computer and use it in GitHub Desktop.
Download ZIP
AI Providers
Raw
AI Providers.md

01. Google Gemini API

  • ==Current: Flash: 5/10 RPM, 20 RPD, 250K TPM | Gemma: 30 RPM + 14.4K RPD + 15K TPM==

Gemini API Rate Limits

Model Category RPM TPM RPD
Gemini 2.5 Flash Text-out models 5 250K 20
Gemini 2.5 Flash Lite Text-out models 10 250K 20
Gemini 2.5 Flash Native Audio Dialog Live API Unlimited 1M Unlimited
Gemini 2.5 Flash TTS Multi-modal models 3 10K 10
Gemini 3 Flash Text-out models 5 250K 20
Gemini 3.1 Flash Lite Text-out models 15 250K 500
Gemini Embedding 1 Other models 100 30K 1K
Gemini Embedding 2 Other models 100 30K 1K
Gemini Robotics ER 1.5 Preview Other models 10 250K 20
Gemma 3 12B Other models 30 15K 14.4K
Gemma 3 1B Other models 30 15K 14.4K
Gemma 3 27B Other models 30 15K 14.4K
Gemma 3 2B Other models 30 15K 14.4K
Gemma 3 4B Other models 30 15K 14.4K
  • requests per minute (RPM)
  • tokens per minute (TPM)
  • requests per day (RPD)

Options:

02. Ollama Cloud Models

  • ==Current: Cloud deployment - 5 Hour Session Usage + 5 Days Weekly Usage (LIMITS)==

Ollama Cloud imposes tiered rate limits to manage capacity and prevent abuse. These vary by plan, with no exact numerical quotas publicly detailed beyond general descriptions.

Designed for light usage like chat and quick model tests. Includes hourly and daily caps, plus per-minute restrictions on rapid API calls.

Free Tier: Ollama Cloud Usage Session usage - Resets in 5 hours Weekly usage - Resets in 5 days

Options:

03. Openrouter Free Models API

  • ==Current: 20 RPM + 1K RPD, :free Models==

OpenRouter uses global, credit-based limits plus a few hard caps. openrouter

Rate limits

  • Requests using :free model variants are capped at 20 requests per minute. openrouter
  • If you have bought less than 10 credits, you can send 50 free-model requests per day; with at least 10 credits purchased, this increases to 1000 per day. openrouter

Credits and key info

  • Each API key has a credit limit (or unlimited), a reset schedule, and remaining credits, all retrievable via GET https://openrouter.ai/api/v1/key. openrouter
  • The key metadata also tracks total, daily, weekly, and monthly usage for both regular and BYOK (bring your own key) traffic, and flags whether the account is on the free tier. openrouter

Other limitations

  • Making multiple accounts or API keys does not bypass rate limits, which are enforced globally per user. openrouter
  • Cloudflare DDoS protection can temporarily block IPs that send traffic far above normal patterns, and a negative credit balance can trigger 402 errors even on free models until credits are added. openrouter

Options:

04. Groq API

  • ==Current: Super-Fast, 30-60 RPM, 250-1K RPD, 100-500k TPD==
MODEL ID RPM RPD TPM TPD ASH ASD
allam-2-7b 30 7K 6K 500K - -
canopylabs/orpheus-arabic-saudi 10 100 1.2K 3.6K - -
canopylabs/orpheus-v1-english 10 100 1.2K 3.6K - -
groq/compound 30 250 70K - - -
groq/compound-mini 30 250 70K - - -
llama-3.1-8b-instant 30 14.4K 6K 500K - -
llama-3.3-70b-versatile 30 1K 12K 100K - -
meta-llama/llama-4-scout-17b-16e-instruct 30 1K 30K 500K - -
meta-llama/llama-prompt-guard-2-22m 30 14.4K 15K 500K - -
meta-llama/llama-prompt-guard-2-86m 30 14.4K 15K 500K - -
moonshotai/kimi-k2-instruct 60 1K 10K 300K - -
moonshotai/kimi-k2-instruct-0905 60 1K 10K 300K - -
openai/gpt-oss-120b 30 1K 8K 200K - -
openai/gpt-oss-20b 30 1K 8K 200K - -
openai/gpt-oss-safeguard-20b 30 1K 8K 200K - -
qwen/qwen3-32b 60 1K 6K 500K - -
whisper-large-v3 20 2K - - 7.2K 28.8K
whisper-large-v3-turbo 20 2K - - 7.2K 28.8K

Groq enforces per-organization limits on requests, tokens, and audio duration, plus response headers to help you throttle when you hit caps. console.groq

What is limited

  • Metrics: requests per minute/day (RPM, RPD), tokens per minute/day (TPM, TPD), and audio seconds per hour/day (ASH, ASD) for speech models. console.groq
  • Cached tokens from prompt caching do not count against token limits. console.groq
  • Limits apply at the organization level, so all API keys and users in that org share the same quotas. console.groq
Header Value Notes
retry-after 2 In seconds
x-ratelimit-limit-requests 14400 Always refers to Requests Per Day (RPD)
x-ratelimit-limit-tokens 18000 Always refers to Tokens Per Minute (TPM)
x-ratelimit-remaining-requests 14370 Always refers to Requests Per Day (RPD)
x-ratelimit-remaining-tokens 17997 Always refers to Tokens Per Minute (TPM)
x-ratelimit-reset-requests 2m59.56s Always refers to Requests Per Day (RPD)
x-ratelimit-reset-tokens 7.66s Always refers to Tokens Per Minute (TPM)
  • If you exceed limits, the API returns HTTP 429 with a retry-after header indicating how many seconds to wait before retrying; this header appears only when a rate limit is hit. console.groq

Options:

05. Cohere API

  • ==Current: 20 RPM (Each Model) + 1000 RPM month (All)==

Chat API rate limits (Markdown table)

Model Trial rate limit Production rate limit
Command A Reasoning 20 req / min Contact sales
Command A Translate 20 req / min Contact sales
Command A Vision 20 req / min Contact sales
Command A 20 req / min 500 req / min
Command R+ 20 req / min 500 req / min
Command R 20 req / min 500 req / min
Command R7B 20 req / min 500 req / min

Other endpoints (Markdown table)

Endpoint Trial rate limit Production rate limit
Embed 2,000 inputs / min 2,000 inputs / min
Embed (Images) 5 inputs / min 400 inputs / min
Rerank 10 req / min 1,000 req / min
Tokenize 100 req / min 2,000 req / min
EmbedJob 5 req / min 50 req / min
Default (other) 500 req / min 500 req / min

Summary of rate limits and limitations

  • There are two API key types: evaluation (trial) keys and production keys; trial keys and production keys on newer chat models are capped at ==1,000 API calls per month==. docs.cohere
  • Chat models generally allow ==20 requests per minute== on trial keys, while mature production models (Command A, R, R+, R7B) allow up to 500 requests per minute; newer Command A variants require contacting sales for production limits. docs.cohere
  • ==Non-chat endpoints have their own per-minute caps==, with Embed text sharing the same limits for trial and production, while others like Rerank and Tokenize get much higher production limits than trial. docs.cohere
  • Any endpoint not explicitly listed falls under a default limit of 500 requests per minute for both trial and production keys, and rate-limit increases require contacting Cohere support. docs.cohere

Options:

06. GitHub Models API

  • ==Current: Low: 15 RPM + 150 RPD | High: 10 RPM + 50 RPD | 8K IN, 4k OUT TPR==
Tier / Model group Metric Copilot Free
Low Requests per minute 15
Low Requests per day 150
Low Tokens per request 8000 in, 4000 out
Low Concurrent requests 5
High Requests per minute 10
High Requests per day 50
High Tokens per request 8000 in, 4000 out
High Concurrent requests 2

Summary of rate limits and limitations

  • Free playground and API usage are limited by requests per minute, requests per day, tokens per request, and concurrent requests, with different tiers for low, high, and embedding models. docs.github
  • Special model groups (Azure OpenAI families, DeepSeek, xAI Grok) have stricter RPM/RPD and concurrency but can offer larger output token caps at higher tiers. docs.github

Options:

07. Mistral API

  • ==Current: 1 RPS, 500k TP-Min, 1KKK TP-Month==

Free plan – feature table

Feature Free plan value Notes
Price €0 / $0 Personal use for life and work.
Flash answers Up to 150 per day Quick “flash” responses.
Web searches Base quota Paid plans allow “Up to 5x Free”.
Think mode Base quota Paid plans: “Up to 30x Free”.
Deep research (preview) Base quota Paid plans: “Up to 5x Free”.
Memories 500 Saved and recallable user memories.
Libraries / storage Limited Higher tiers raise to 15–30 GB.
Document uploads Base quota Paid: “Up to 20x Free”.
Image generation Base quota Paid: “Up to 40x Free”.
Code interpreter Base quota Paid: “Up to 5x Free”.
Projects Unlimited Can group chats into projects.
Connectors directory Full access Access to connectors directory.
Custom MCP connectors Not included Marked as “Custom” only on higher tiers.
Voice / canvas / agents Limited or not listed Described as “Custom” mainly for paid tiers.
Customer support Help center only No chat/email support on Free.

Rate limits and other limitations (Free only)

  • The Free plan gives access to Mistral’s SOTA models in Le Chat but applies hard caps on usage-based features such as ==flash answers (150/day)==, web searches, extended “think” reasoning, deep research, document uploads, image generation, and code interpreter runs, all at a baseline level that paid plans multiply (for example “Up to 40x Free” for images).
  • You can store up to 500 memories, have unlimited projects, and full access to the connectors directory, but storage (libraries) is limited and advanced enterprise features like domain verification, audit logs, SAML SSO, and white-labeling are not available. mistral

Options:

08. Cerebras API

  • ==Current: 30 RPM + 900 RPH + 14k RPD + 1000K TPD==

Free tier rate‑limit table (Markdown)

Model TPM TPH TPD RPM RPH RPD
gpt-oss-120b 64,000 1,000,000 1,000,000 30 900 14,400
llama3.1-8b 64,000 1,000,000 1,000,000 30 900 14,400
qwen-3-235b-a22b-instruct-2507 64,000 1,000,000 1,000,000 30 900 14,400
zai-glm-4.7 64,000 1,000,000 1,000,000 10 100 100
  • Due to high demand on zai-glm-4.7 and qwen-3-235b-a22b-instruct-2507, we’ve temporarily reduced free-tier rate limits. We’re working to bring it back as quickly as possible and appreciate your understanding.

Summary of free‑tier rate limits and other limitations

  • Cerebras measures limits on both requests (RPM, RPH, RPD) and tokens (TPM, TPH, TPD), and whichever limit you hit first will throttle further calls. inference-docs.cerebras

  • Free‑tier org limits for most models are ==30 RPM, 900 RPH, 14.4K RPD and 60K TPM, 1M TPH, 1M TPD==, while the heavier zai-glm-4.7 model is restricted to ==10 RPM, 100 RPH, and 100 RPD== with the same token caps. inference-docs.cerebras

  • Every response includes rate‑limit headers (limit, remaining, and reset for requests/day and tokens/minute); exceeding limits returns HTTP 429, and increasing quotas requires contacting Cerebras. inference-docs.cerebras

  • TPM (tokens/min); TPH (tokens/hour); TPD (tokens/day)

  • RPM (requests/min); RPH (requests/hour); RPD (requests/day)

Options:

09. Cloudflare API

  • ==Current: 1K RPM; 100K RPD; Free Workers: 10k Neuros/day==
Feature Workers Free Workers Paid
Request 100,000 requests/day
1000 requests/min		 No limit
Worker memory 128 MB 128 MB
CPU time 10 ms 5 min HTTP request
15 min Cron Trigger
Duration No limit No limit for Workers.
15 min duration limit for
Cron Triggers,
Durable Object Alarms and 
Queue Consumers

Cloudflare Workers Free gives you limited daily traffic and tighter per-request resources compared to paid plans. developers.cloudflare

10. Huggingface Inference API

  • ==Current: $0.10 / month usage==

Free plan rate‑limit table (Markdown)

Plan API requests / 5 min Resolver requests / 5 min Pages requests / 5 min
Free user 1,000 * 5,000 * 200 *
    • Anonymous and Free user limits may change over time depending on platform health. huggingface

Summary – rate limits and other limitations (Free only)

  • Hugging Face defines separate buckets for Hub API, Resolvers (file downloads via /resolve/), and Pages, each with its own quota over a 5‑minute fixed window. huggingface
  • A Free user can make about 1,000 API calls, 5,000 resolver calls, and 200 page views per 5‑minute window, but these values are explicitly marked as subject to change for Free and anonymous users. huggingface
  • When you hit a limit, the Hub returns HTTP 429 with standardized RateLimit and RateLimit-Policy headers that describe remaining quota and reset time, and the recommended mitigation is to authenticate with HF_TOKEN, spread requests over time, or prefer Resolver endpoints where possible. huggingface

Options:

11. Nvidia NIM API

  • ==Current: UNLIMITED: 40 RPM==

NVIDIA does not actually publish concrete free‑tier model limits for NIM in this thread, so there is no detailed table beyond the one mentioned value. forums.developer.nvidia

Minimal “table” from the discussion

Aspect Trial / free experience detail
Published per‑model limits Not published
Example mentioned by the user 40 requests per minute (from NVIDIA API catalog trial)
Token / context window limits Not disclosed

Summary – rate limits and other limitations (trial / free)

  • An NVIDIA staff moderator confirms that they do not publish specific limits for each NIM model, even for the trial / catalog experience. forums.developer.nvidia
  • The thread references a 40 requests per minute cap in the catalog trial, but NVIDIA does not confirm or extend this with token or context‑window numbers. forums.developer.nvidia

Options:

12. Vercel API

  • ==Current: $5/- Per Month==

Vercel AI Gateway’s free tier is credit-based rather than having explicit RPM/TPM quotas. vercel

Plan Monthly credit When free ends
AI Gateway $5 When you purchase any AI Credits

Summary – rate limits and other limitations (free only)

  • Every team gets $5 of AI Gateway Credits every 30 days after making the first AI Gateway request; these credits can be used on any provider/model in the catalog at standard, no‑markup rates. vercel

Options:

13. Iflow API

  • ==Current: UNLIMITED: 1 RPS — till 429 error==
Limit type Value / behavior
Concurrent requests 1 request per user at a time (global concurrency limit)
Over-limit behavior Additional requests return HTTP 429
Pricing Free to use
Streaming requests Tokens released immediately after active cancellation
Non-streaming Model continues running;
tokens released only after completion, even if canceled

Summary – rate limits and limitations (free only)

  • The platform allows only one active request per user at any time; any extra concurrent request receives a 429 error. platform.iflow
  • Usage is currently completely free, but users are asked to avoid unnecessary high concurrency to protect shared resources. platform.iflow
  • The list of Models: https://platform.iflow.cn/en/models

Options:

14. Perplexity API

  • ==Current: $0/- Per Month==

Perplexity Pro affects billing/credits, not API rate limits directly. perplexity

Plan Monthly API credits Credit refresh timing Notes
Perplexity Pro $0 1st day of each month Auto-applied

Summary – limits and other conditions (Perplexity Pro)

  • Pro subscribers automatically receive $0 in API credits on the first day of each month, and new subscribers see the credit appear within about 10–20 minutes after subscribing.

Options:

15. AI ML API

  • ==Current: 10 RPH, 50K Credits-Per-Day, $0.025/- Per Day==
Plan type Requests Models available Credit / cost limits Notable exclusions
Free Unverified 10 requests per hour Gemma 3 4b, Gemma 3 12b, Gemma 3n 4b No explicit daily credit cap mentioned All other models
Free Verified 10 req/hour on FREE models; 10 req/day on cheap paid models Any model costing < 50,000 credits per request 50,000 credits per day; model cost ≤ $0.025 or 50,000 credits Audio, video, and most image models (typically > 50,000 credits)

Summary of rate limits and other limitations

  • On Free Unverified, new users get 10 requests per hour and can only use three Gemma 3 family models (4b, 12b, 3n‑4b), with no broader model access. help.aimlapi
  • On Free Verified, after adding a payment method, you can use any model that costs under 50,000 credits per request, but you are limited to 10 requests per hour on FREE‑label models, 10 requests per day on low‑cost paid models, and a total of 50,000 credits per day. help.aimlapi
  • Even when verified, audio, video, and most image models are not free, because a single call to those models usually exceeds the 50,000‑credit threshold. help.aimlapi

Options:

16. Opencode Zen API

  • ==Current: UNLIMITED: 1 RPS - Free 2 Models==

Zen does not have a classic “free tier” with rate limits; instead, some models are priced at $0 per 1M tokens, and the platform lets you cap spend with monthly limits. opencode

Model Input price / 1M tokens Output price / 1M tokens Cached read / 1M tokens Cached write / 1M tokens Notes
Big Pickle Free Free Free – Free during limited beta
GPT 5 Nano Free Free Free – Always-free listed model
MiMo V2 Flash Free Free Free Free – Free during limited beta
Nemotron 3 Super Free Free Free Free – Free during limited beta
MiniMax M2.5 Free Free Free Free – Free during limited beta

Summary – limits and other conditions for “free” usage

  • Zen is pay‑as‑you‑go; there is no fixed free allowance, but models like Big Pickle and GPT 5 Nano are currently priced at $0 for input, output, and cached reads, so they effectively function as free to use while available. opencode

Options:

17. Pollination AI API

  • ==Current: UNLIMITED: 1 RP15S - Free 2 Models==

Pollinations.ai provides a free, open-source API alternative compatible with OpenAI-style endpoints for text, image, and audio generation. pollinations

No signup or API keys are needed for basic use, prioritizing privacy with zero data storage. github

Key Features

  • Supports models like GPT-5, Claude, Gemini for text; Flux, GPT Image for images; and OpenAI-audio for speech. pollinations-ai
  • Fully open-source on GitHub, used in 500+ community projects including bots and apps.
  • Earn daily "Pollen Credits" via contributions for advanced access; basic is unlimited and free. pollinations

Usage Examples

Access via simple URLs:

  • Images: https://pollinations.ai/p/your prompt here github
  • Text: https://text.pollinations.ai/your prompt github
  • Audio: Append ?model=openai-audio&voice=nova to text URL. github

Integrate in code, like Python requests for images or React hooks for apps. github

OpenAI Compatibility

Offers OpenAI-compatible text models listable via API; proxy-like access to premium models without direct OpenAI keys. enter.pollinations

Options:

18. Long Cat

  • ==Current: UNLIMITED: 1 RPS — till 429 error==

LongCat is a free LLM API platform compatible with OpenAI and Anthropic formats, offering daily token quotas, several in‑house models, and simple HTTP/SDK integration, with token-output limits but no explicit global QPS limits documented. longcat

Overall Summary

  • Supported models include ==LongCat-Flash-Chat, LongCat-Flash-Thinking, LongCat-Flash-Thinking-2601, and LongCat-Flash-Lite==, all available via both API formats, with different performance profiles (general chat, deep thinking, MoE, upgraded deep thinking). longcat

  • Every account gets free daily token quotas: ==500,000 tokens/day== for the Flash-Chat / Flash-Thinking / Flash-Thinking-2601 models, and ==50,000,000 tokens/day== for LongCat-Flash-Lite; quotas reset at Beijing midnight and do not roll over. longcat

  • You can apply via the Usage page to raise the free quota for the first three models up to ==5,000,000 tokens/day== (Flash-Lite is excluded), or email the team with an API key for more quota; both input and output tokens count, and streaming vs non-streaming consume quota equally. longcat

  • ==Usage can be checked in real time== on the Usage page, and at the moment the platform is in public beta and does not support paid quota purchases. longcat

  • If you exceed platform rate limits, the ==API returns HTTP 429== with an error payload containing code: "rate_limit_exceeded", type: "rate_limit_error", and a retry_after field (in seconds, example value 60) telling you when to retry. longcat

Options:

19. Google Gemini CLI

  • ==60 RPM + 1k RPD; 120 RPM + 1.5k RPD==

For the paid Gemini Code Assist subscriptions (what they effectively treat as the “Gemini Pro” style plans), the CLI lists higher fixed quotas. geminicli

Plan Requests / user / day Requests / user / minute Models used
Code Assist Standard edition 1500 120 Gemini family (auto-chosen)

Summary – Gemini “Pro”/paid limits and behavior

  • Upgrading from the free Google-login tier to Gemini Code Assist Standard raises your quota to 1500 model requests per user per day with 120 requests per user per minute, spread across the Gemini family as selected by the CLI. geminicli
  • The Enterprise edition increases only the daily cap (to 2000 requests per user per day) while keeping the same 120 requests per minute limit. geminicli
  • These paid tiers give fixed-price, predictable quotas; if you still need more or want to avoid hitting daily caps, you must switch to pay‑as‑you‑go via a Gemini API key or Vertex AI, where limits are governed by API pricing tiers and Vertex quotas rather than these fixed per-user request counts. geminicli

20. Iflow CLI

  • ==Current: 1 RPS — Till 429 Error Threshold==

The page documents configuration knobs, not hosted rate limits; the only “limits” you can set or hit are things like session turns and token budgets. platform.iflow

Setting / limit Default value What it does
Current RPS guidance ~1 RPS until 429 Recommended request rate
maxSessionTurns -1 Unlimited turns per chat session
tokensLimit 128000 Maximum context window length
compressionTokenThreshold 0.8 Auto-compress

Summary for free usage

  • iFlow CLI itself does not impose a strict free-tier quota; instead, it recommends keeping around 1 request per second to avoid 429 errors from whichever backend you configure. platform.iflow

21. Kilo Code CLI

  • ==Current: Varies by provider: ~60 RPM + ~2,000 RPD on some free plans==

Kilo Code itself does not impose numeric rate limits; it relies on free quotas from external providers and on models that are priced at $0. kilo

Category Model / Provider Cost in Kilo Code
OpenRouter free models Qwen3 Coder (free) Free via OpenRouter
OpenRouter free models Z.AI GLM‑4.5 Air (free) Free via OpenRouter
OpenRouter free models DeepSeek R1 0528 (free) Free via OpenRouter
OpenRouter free models MoonshotAI Kimi K2 (free) Free via OpenRouter

22. Mistral Vide CLI

  • ==Current: 1 RPS + 500K TPM - 1B TPM Monthly==

Mistral’s docs describe how limits work but do not publish concrete free‑tier numbers on this page. docs.mistral

Aspect Free API tier behavior
Availability Yes, a dedicated free API tier
Purpose Trying and exploring the API, not production workloads
Limit types Requests per second (RPS), tokens per minute/month
Scope Limits applied at workspace level
Configuration/visibility Exact limits shown only in AI Studio “limits” page per workspace
Upgrades Higher tiers provide higher limits; contact support to increase

23. Opencode CLI

  • ==Current: UNLIMITED: 1 RPS - Free 2 Models==

Zen does not have a classic “free tier” with rate limits; instead, some models are priced at $0 per 1M tokens, and the platform lets you cap spend with monthly limits. opencode

Model Input price / 1M tokens Output price / 1M tokens Cached read / 1M tokens Cached write / 1M tokens Notes
Big Pickle Free Free Free – Free during limited beta
GPT 5 Nano Free Free Free – Always-free listed model

Summary – limits and other conditions for “free” usage

  • Zen is pay‑as‑you‑go; there is no fixed free allowance, but models like Big Pickle and GPT 5 Nano are currently priced at $0 for input, output, and cached reads, so they effectively function as free to use while available. opencode

24. Qwen CLI

  • ==Current: 60 RPM + 2000 RPD - No Token Limits==
Free option Requests per day Requests per minute Token limits
Qwen OAuth 2,000 60 No token counting

Summary of rate limits and other limitations

  • With Qwen OAuth, individual users get 2,000 requests per day and 60 requests per minute, advertised as completely free with no explicit token limit tracking on your side. npmjs
  • Regional free tiers add more options: ModelScope offers 2,000 free API calls per day for Mainland China, and OpenRouter offers up to 1,000 free calls per day internationally, each subject to their own backend quotas and terms. npmjs
  • Qwen Code warns that a single workflow cycle may trigger multiple API calls, so heavy commands can consume your free request budget faster than expected even though there is no per-call token cap. npmjs

25. GitHub Copilot CLI

  • ==Current: 50 request per month.==
Plan Monthly price Premium requests (chat, agents, reviews, CLI) Code completions Models access
Free $0 50 per month 2,000 per month Haiku 4.5, GPT‑4.1, GPT‑5 mini, and more

Summary – rate limits and other limitations (Free only)

  • The Free Copilot plan includes 50 premium requests per month, which are consumed by ==chat, agent mode, code review, coding agents, and Copilot CLI features==; once used up, those premium features stop working until the next month. github
  • You also get 2,000 code completions per month in supported editors; after hitting that cap, inline suggestions are disabled until your quota resets. github
  • Free users still have access to multiple modern models (such as Claude Haiku 4.5, GPT‑4.1, GPT‑5 mini) but see rate limiting and slower response times during high usage, and they cannot buy extra premium requests or access enterprise governance features. github

Models in CLI

  1. Gemini CLI
    1. Gemini 3 - gemini-3.1-pro-preview, gemini-3-flash-preview
    2. Gemini 2.5 - gemini-2.5-pro, gemini-2.5-flash, gemini-2.5-flash-lite
  2. OpenAI Codex
    1. gpt-5.4
    2. gpt-5.3-codex
    3. gpt-5.2-codex
    4. gpt-5.2
    5. gpt-5.1-codex-max
    6. gpt-5.1-codex-mini
  3. IFlow CLI
    1. GLM-4.7
    2. iFlow-ROME-30BA3B
    3. DeepSeek-V3.2
    4. GLM-5
    5. Qwen3-Coder-Plus
    6. Kimi-K2-Thinking
    7. MiniMax-M2.5
    8. Kimi-K2.5
    9. Kimi-K2-0905
  4. Kilocode CLI
    1. Kilo Auto Free
    2. CoreThink
    3. Giga Potato
    4. Giga Potato Thinking
    5. MiniMax M2.5
    6. Trinity Large Preview
    7. Stepfun Step 3.5 Flash
    8. Nemotron 3 Super
    9. Groq Code Fast 1 Optimied
  5. Mistral Vibe CLI
    1. Devstral 2
    2. Devstral Small
    3. Local
  6. OpenCode Zen CLI
    1. Big Pickle
    2. GPT 5 Nano
    3. MiniMax M2.5 Free
    4. Nemotron 3 Super Free
    5. MiMo V2 Flash Free
    6. Trinity Large Preview Free
  7. Qwen CLI
    1. Latest Qwen Coder model - Qwen 3.5 Plus
  8. GitHub Copilot CLI
    1. Claude Haiku 4.5
    2. GPT-5 mini
    3. GPT-4.1
  9. Cline CLI
    1. arcee-ai/trinity-large-preview:free
    2. openrouter/free

Updated on February 6th, 2026 16:55:44

Opencode CLI - AI Providers Connect

OpenCode connects to over 75 LLM providers via the AI SDK and Models.dev, including OpenAI, Anthropic, ==Google==, Together AI, ==Mistral AI==, ==Groq==, Meta, ==Cohere==, AWS Bedrock, Azure OpenAI, DeepSeek, ==Perplexity API==, ==Hugging Face==, Replicate, Fireworks AI, Lepton AI, Novita AI, AI21 Labs, Aleph Alpha, Bamboo AI, ==Cerebras==, ==Cloudflare Workers AI==, EleutherAI, Falcon, GLM (Zhipu), Google Vertex AI, xAI (Grok), IBM watsonx, Inflection, LightOn, Moonshot AI, ==NVIDIA Nemo==, ==Ollama==, ==OpenRouter==, PaddlePaddle, ==Qwen (Alibaba)==, Stability AI, Yi, and many additional OpenAI-compatible endpoints accessible through custom configurations or OpenCode Zen. opencode

Kilocode CLI - AI Providers Connect

Kilo Code supports over 30 AI providers, including cloud options like Anthropic, OpenAI, ==Google Gemini==, DeepSeek, ==Mistral==, and Zhipu AI; local and self-hosted setups such as ==Ollama==, LM Studio, and any OpenAI-compatible endpoints; AI gateways and routers including ==OpenRouter==, Glama, Requesty, and its own Kilo Gateway; plus enterprise choices like AWS Bedrock and Google Vertex AI. kilo

LLMLite

LiteLLM (often called llmlite in shorthand contexts) is a Python library and proxy server that unifies access to 100+ LLM providers through a single OpenAI-compatible format, proxying calls to endpoints like Bedrock, Azure OpenAI, Vertex AI, ==Google AI Studio==, Anthropic, AWS SageMaker, ==Cohere==, ==Hugging Face==, Replicate, ==Groq==, ==Mistral AI==, ==Cloudflare Workers AI==, DeepInfra, ==Perplexity AI==, DeepSeek, Anyscale, IBM watsonx.ai, Voyage AI, FriendliAI, Galadriel, ==OpenRouter==, ==Ollama==, vLLM, Together AI, Fireworks AI, Novita AI, ==Cerebras==, Baseten, AI21, Aleph Alpha, NLP Cloud, Petals, Dashscope (Qwen), Sarvam.ai, OVHCloud AI Endpoints, ==CometAPI==, CompactifAI, DataRobot, ElevenLabs, Fal AI, Featherless AI, GradientAI, Helicone, Hyperbolic, Infinity, Jina AI, Lambda AI, Llamafile, LM Studio, LlamaGate, Manus, Meta Llama, Milvus, MiniMax, Moonshot AI, Morph, Nebius AI Studio, Novita AI, Nscale, ==NVIDIA NIM==, Oracle OCI, Poe, Predibase, PublicAI, Recraft, RunwayML, SambaNova, SAP Generative AI Hub, Scaleway, Stability AI, Synthetic, Snowflake, Topaz, Triton Inference Server, ==Vercel AI Gateway==, Volcano Engine, Weights & Biases Inference, Xiaomi MiMo, Xinference, Zhipu AI (Z.AI), and many more via custom or OpenAI-compatible setups. docs.litellm

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment