What is the cheapest good AI model for everyday coding and automation?
Last verified 2026-10-15Paid placement never changes our picks.
Short answer
As of October 2026, start with the OpenAI API if you want a first-party small model at a low listed rate, the Gemini API if you want a free tier to test on, and DeepSeek if you want low off-peak rates. The catch: cheapest per token is not cheapest per task, so price your own input and output mix.
Top picks
A first-party small model for high-volume coding and automation
- Free tier
- No free tier
- Starting price
- See pricing page
- Platforms
- API
- Last verified
- 2026-10-10
GPT-6 Luna lists $0.10 in and $0.50 out per million tokens; prompts over 272K tokens cost more.
Testing on a free API tier before paying
- Free tier
- Free tokens on certain models via AI Studio; free-tier content may be used to i…
- Starting price
- See pricing page
- Platforms
- API, AI Studio
- Last verified
- 2026-10-12
Free-tier content may be used to improve Google products. Paid 3.5 Flash-Lite is $0.30 in and $2.50 out per million.
Low off-peak rates with tool calls and code completion
- Free tier
- App stated as free, no ads or in-app purchases; web and desktop not stated
- Starting price
- Free
- Platforms
- API
- Last verified
- 2026-10-10
V4.1 Flash is $0.15 in and $0.60 out per million off-peak; rates double at weekday peak hours (UTC).
Short-prompt, high-volume work at the same listed rate
- Free tier
- No free tier
- Starting price
- See pricing page
- Platforms
- API
- Last verified
- 2026-10-12
Built by Anthropic, the maker of the model that drafted this page; held to the same standard. Rate applies under 100K-token prompts.
Open-weight model family with cheap cached input
- Free tier
- No free tier
- Starting price
- See pricing page
- Platforms
- API
- Last verified
- 2026-10-12
GLM-5.3-Flash is $0.15 in and $0.50 out per million; weights licence not read. Also hosted by third parties.
Apache 2.0 weights you can also call by API
- Free tier
- Limited messages and web searches, limited coding sessions, basic image generat…
- Starting price
- From $14.99
- Platforms
- API
- Last verified
- 2026-10-10
Mistral Small 4 is $0.15 in and $0.60 out per million tokens and lists an Apache 2.0 licence.
Running open models locally with no per-token cost
- Free tier
- Run models locally at no cost; starter cloud credits; 1 concurrent cloud request
- Starting price
- From $20
- Platforms
- Windows, macOS, Linux
- Last verified
- 2026-10-10
Local models are free on every plan; speed depends on your GPU and memory. Paid plans add cloud usage.
Prices come from each tool's fact block. "See pricing page" means we have not confirmed the number on the official pricing page in the last 30 days, so we do not show one.
How to choose
- If you pay per token → price your own mix of input and output (output costs about 4 to 8 times input on most models here, so chatty coding agents cost more than classifiers); use caching where offered.
- If you want a first-party small model → OpenAI's GPT-6 Luna or Anthropic's Claude Haiku 5.5 (the same listed short-prompt rate; both charge more once a prompt passes a length threshold).
- If you want to test for free → the Gemini API (a free tier for Flash-Lite, but free-tier content may be used to improve Google products; the paid tier says it is not).
- If you want open weights or self-hosting → Mistral Small 4 (Apache 2.0 listed), GLM through Z.ai, or Ollama on your own PC (no per-token cost, but your hardware limits speed); check each model's licence first.
- If the job is hard multi-file coding → test a mid-tier model on one real task first (vendors position these small models for high-volume work); see /ask/top-vs-mid-tier-ai-model.
How we checked
- 7 of 7 picks have a Toolsfine fact block.
- 7 of 7 picks had their pricing checked on the official site in the last 30 days.
- 1 of 1 models also suggested OpenAI API.Panel: Claude (2026-10-12).
- Verified on 2026-10-15.
Listed in other directories
| Tool | theresanaiforthat.com | futurepedia.io | toolify.ai |
|---|---|---|---|
| OpenAI API (GPT-6 Luna) | Not checked | Not checked | Not checked |
| Gemini API (Flash-Lite) | Not checked | Not checked | Not checked |
| DeepSeek | Not checked | Listed | Not checked |
| Claude API (Haiku 5.5) | Not checked | Not checked | Not checked |
| Z.ai (GLM) | Not checked | Not checked | Not checked |
| Mistral (Small 4) | Not checked | Not checked | Not checked |
| Ollama | Not checked | Listed | Not checked |
"Not checked" means we could not open that directory's page; it does not mean the tool is missing there.
Where sources disagree
- Fairness disclosure: Claude is built by Anthropic, the maker of the model that drafted this page; held to the same standard. Anthropic rows use the same fields and plain wording as the others, and ranks follow the reader's situation, not preference. Claude Haiku 5.5 and GPT-6 Luna list the same short-prompt rate; the tie is broken against Anthropic's model.
- Tiered and time-based prices: The low rates have conditions: Claude Haiku above 100K prompt tokens and GPT-6 Luna above 272K cost more, and DeepSeek doubles at weekday peak hours (UTC).
- Promotions and model names: Gemini 3.8 Flash has a promotional price that doubles on January 1, 2027. Some OpenAI pages still list the older GPT-5.6 Luna at higher rates; the rates here are for GPT-6 Luna.
- Open-weight licences: Only Mistral's Apache 2.0 statement was read. Whether DeepSeek and GLM weights are released, and under what licence, was not verified.
Also mentioned elsewhere
Other sources name these tools for this job. They are not in our picks because we have not checked them yet.
- Qwen on Alibaba Cloud Model Studio · Official siteWidely used open-weight model family; pricing page not checked.
- OpenRouter · Official siteRoutes one API key to many hosted models; rates not compared here.
- Fireworks · Official siteHost for open models; per-model rates were not on the page text we read.
- Groq · Official siteFast inference host; its pricing page redirected to the homepage on our check.
Know a tool we should check? Submit it →
Related answers
- Is the top-tier AI model worth paying for over the mid-tier one?
- Which AI assistant connects to the apps I already use?
- Which AI tool helps turn saved notes and links into action?
- Which AI voice assistant is good for talking things through?
- Which AI agents can do computer tasks while keeping my files private?
Changelog
- 2026-10-15 · Picks, prices and sources re-checked
- 2026-10-15 · Page updated
- 2026-10-15 · First published
Last verified 2026-10-15 · next re-check 2027-01-13. Report an error
Researched with AI assistance and checked by our editors. AI can make mistakes and vendors change plans often — confirm prices and terms on the vendor's site before you buy.
Paid placement never changes our picks. Data: JSON