Every few weeks a new model makes the news, and the first question many people ask is whether they can try it for free. For the big releases of September and October 2026 the honest answer is mostly no. The models you can use for free through an API are a step behind, but they are good enough for most everyday questions.
The new models and their free options
| Model | Maker | Free API? | Images | Closest free alternative |
|---|---|---|---|---|
| Gemini 4 Argon | No. Not on the Gemini API free tier | Yes | Gemini 3.8 Flash, same key, free | |
| GPT-6 Luna, Sol, Astra | OpenAI | No. Paid API only | Yes | GPT-OSS 120B on Groq |
| Claude Opus, Sonnet and Haiku 5.5 | Anthropic | No. Paid API only | Yes | Gemini 3.8 Flash or Groq |
| Kimi K3.1 Pro | Moonshot | No free option found; no :free version on OpenRouter |
Yes | Qwen 3.8 27B on Groq |
| DeepSeek V4.1 Flash | DeepSeek | No free tier stated; one of the cheapest paid options | Yes | GPT-OSS 120B or Qwen 3.8 27B on Groq |
| Mistral Large 4 | Mistral | Possibly, through Mistral’s Free mode; coverage is not stated | Yes | Mistral Free mode |
| Nemotron 3.5 Super VL | NVIDIA | Yes, as a :free model on OpenRouter (50 requests a day) |
Yes | — |
“No” here means no free option on the maker’s own API and no free variant on OpenRouter, checked on 29 October 2026. Consumer chat apps may still let you try a model in a browser; that is a different thing from an API key you can use in other apps.
What is actually free right now
These models are free to call through an API and are good choices for questions on text and images:
- Gemini 3.8 Flash, 3.6 Flash and 3.5 Flash-Lite on the Gemini API free tier. All read images. See how to get a free Gemini API key.
- GPT-OSS 120B and Qwen 3.8 27B on Groq’s free plan, with up to 1,000 requests and 200,000 tokens a day per model. Qwen reads images. See how to get a free Groq API key.
- Gemma 4 31B and Gemma 4 26B as
:freemodels on OpenRouter, 50 requests a day. See how to get a free OpenRouter key. - Gemma 4 26B on Cloudflare Workers AI, within 10,000 free Neurons a day.
The full comparison of free tiers is in best free AI APIs in 2026.
If you want the newest models anyway
The new flagships are paid per token, and some are cheap. Official prices per million input and output tokens, as listed on 29 October 2026:
| Model | Input | Output |
|---|---|---|
| GPT-6 Luna | $0.10 | $0.50 |
| Claude Haiku 5.5 (prompts up to 100,000 tokens) | $0.10 | $0.50 |
| DeepSeek V4.1 Flash | $0.30 | $1.20 |
| Gemini 4 Argon | $2.00 | $10.00 |
DeepSeek charges half outside its peak hours, and Gemini 4 Argon’s price is an introductory one until 30 November 2026, after which Google lists $4 and $20. Prices change often; check the maker’s pricing page before you rely on them.
Using these models in LuciQA
LuciQA is a free desktop app that answers questions about what is on your screen, with the provider and model you choose. It already offers the free models above: Gemini 3.8 Flash, Groq’s GPT-OSS and Qwen models, and the free Gemma 4 models on OpenRouter and Cloudflare. If you pay for API use, it also offers Gemini 4 Argon on a billed Gemini key, GPT-6 Luna, GPT-6.1 Sol and GPT-6 Astra on an OpenAI key, and the Claude 5.5 models on an Anthropic key.
With LuciQA Pro, several providers can answer the same question side by side, so a free Gemini answer and a free Groq answer can check each other. This page is updated as new models arrive.