Free AI models for coding, every day

Five models are free to use in Clixad's terminal coding agent every day, with no API key and no card. They share one allowance of 60,000 credits of usage per account per day, and this page shows what that buys on each of them.

  • No API key
  • No monthly subscription
  • Free models every day
npm install -g clixad

Node 20 or newer. Then clixad login signs you in with GitHub, and clixad in a project opens the agent.

5models free every day
60,000credits a day, shared by all five
100free turns a day on GPT-5 Nano
00:00 UTCwhen the allowance resets

The five free models

A turn is 20,000 input and 1,500 output tokens, the size the /model picker estimates with. Context windows as a Clixad session gets them, which for Claude Haiku 5.5 is less than OpenRouter lists (below). Recomputed from packages/pricing by a test.
ModelVendorContext window, tokensCredits per turnFree turns a day
GPT-5 NanoOpenAI400,000240250
Gemini 2.5 Flash-LiteGoogle1,048,576390153
Claude Haiku 5.5Anthropic24,000413145
DeepSeek V4 FlashDeepSeek1,048,576484123
Gemini 3.5 Flash-LiteGoogle1,048,5761,46341

GPT-5 Nano is the cheapest model in the whole catalogue, so the allowance goes furthest on it. DeepSeek V4 Flash is the model a new account starts on. Claude Haiku 5.5 is the one Claude model in the set. Gemini 3.5 Flash-Lite is the dearest of the five, about six times GPT-5 Nano per turn, which is why the same allowance buys 41 turns on it. Three of the five have a context window of a million tokens or more.

Claude Haiku 5.5 takes a million tokens as well, but every provider of it charges five times the price for a prompt of 100,000 tokens or more. Clixad never sends one, so a session on it gets the 24,000 in the table, and older messages are summarised once a conversation passes three quarters of that.

One allowance, five models

The 60,000 credits are one pool, not a quota per model. A turn on Gemini 3.5 Flash-Lite draws 1,463 credits from it and a turn on GPT-5 Nano draws 240, so how far the day goes depends on which rows you use. You can switch at any point with /model, where each free row shows what it has left, for example free: ~123 turns left.

Free usage is not a credit grant. Nothing is added to your balance for it and nothing is taken away, so the 25,000 credits a new account starts with stay where they are while you work on the free rows.

When the free turns end

Two limits can end them. One is yours: 60,000 credits of usage a day. The other is a daily budget shared by everybody using Clixad, so on a busy day the free turns can end before you reach your own limit. Both reset at 00:00 UTC.

When either is reached, the five models cost credits like the other eight, at the per-turn prices in the table. You can spend the signup bonus, earn credits by completing an offer in a browser tab, credited on completion, or buy a one-time pack. A screenout, where a survey ends part way with nothing or a small bonus, is the most common outcome of starting a survey.

The terminal also shows advertising while a model call runs, one line marked Ad and up to two more places marked the same way. Nothing you type takes part in choosing them, and CLIXAD_SPONSOR=0 stops every request for them.

Other ways to code with a free model

A free model does not have to come from us. Run one locally: Aider says it can connect to "almost any LLM, including local models", and OpenCode lists "75+ LLM providers through Models.dev, including local models". A local model costs no tokens and needs hardware that can run it. Google's own free tier: Gemini CLI with a personal Google account allows 1,000 model requests a day on Gemini Flash models. OpenCode also says "Free models included" on its own site. How to vibe code for free prices each route.

FAQ

Which AI models are free for coding in Clixad?

Five: GPT-5 Nano, Gemini 2.5 Flash-Lite, Claude Haiku 5.5, DeepSeek V4 Flash and Gemini 3.5 Flash-Lite. They share one allowance of 60,000 credits of usage per account per day, with no API key and no card.

How many free turns do I get a day?

It depends on the model, because the allowance is counted in credits. For a turn of 20,000 input and 1,500 output tokens it is about 250 turns on GPT-5 Nano, 153 on Gemini 2.5 Flash-Lite, 145 on Claude Haiku 5.5, 123 on DeepSeek V4 Flash and 41 on Gemini 3.5 Flash-Lite.

When does the free allowance reset?

At 00:00 UTC every day. The allowance also sits inside a daily budget shared by everybody, so on a busy day the free turns can end before your own limit does.

What happens when the free turns run out?

The same models cost credits until the next reset, at the prices the /model picker shows. You can spend the 25,000-credit signup bonus, earn credits on an offer wall in your browser, or buy a one-time pack.

Sources

  1. The free set, the allowance and the turn size: FREE_MODEL_IDS, freeTierDailyCreditsPerUser, freeTierGlobalDailyBudgetUsd and FREE_TURN_USAGE in packages/pricing. Context windows: servedContextTokens in the same package, which is the model's own window as OpenRouter's model list gave it on 27 September 2026, and for Claude Haiku 5.5 the window kept under its providers' price for long prompts, read from OpenRouter's endpoints on 17 October 2026. That free usage adds and removes nothing: sections 6 and 9 of our terms.
  2. Every Clixad figure: packages/pricing (creditsForUsage, signupBonusCredits, freeTierDailyCreditsPerUser, FREE_MODEL_IDS, maxAdRewardUsdPerDay, CREDIT_PACKS), over the model catalogue read from OpenRouter on 27 September 2026, recomputed by the test beside this page. A turn is 20,000 input and 1,500 output tokens, FREE_TURN_USAGE, the size the /model picker estimates with.
  3. Aider, aider.chat, and OpenCode, opencode.ai, for the quoted wording on local and included models. Google, Gemini CLI: Quotas and pricing, for the 1,000 model requests a day. All read 29 September 2026.