Ollama Free Tier Cloud Models Comparison
The list below is as of 19 October 2026.
Free usage credits can be used with the following cloud models:
New: beam is Reflection's first open-weight model (Apache 2.0). Ollama announced it on 16 October 2026 and says it is free for a limited time only, so don't count on it staying free.
Model Comparision and Capabilities¶
Here's a comparison based on the context and usage figures from the sources I found. Sources disagree on some numbers, and the Ollama model pages are the authoritative reference.
| Model | Size | Context | Usage level | Capabilities | Best for |
|---|---|---|---|---|---|
| gpt-oss:20b | 20B | 128K | Low | Text, Tools, Thinking | Quick questions, everyday coding, saving credits |
| nemotron-3-nano:30b | 30B (3B active) | Up to 1M | Low | Text, Tools, Thinking | Long files, many-file context, fast agent tasks |
| gemma4:31b | 31B | 256K | Medium | Text, Vision, Audio, Tools, Thinking | Coding and general chat, image input |
| gpt-oss:120b | 120B | 128K | Medium | Text, Tools, Thinking | Harder reasoning and coding |
| nemotron-3-super | 120B (12B active) | 256K | Medium | Text, Tools, Thinking | Agentic tasks and multilingual work |
| beam | ~501B (23B active) | 1M | Not listed | Text, Tools, Thinking | Coding agents, software engineering, long-horizon tasks |
| nemotron-3-ultra | ~550B (55B active) | 256K | High | Text, Tools, Thinking | Heavy, long-running agent work |
Simple recommendation for Continue:
- Start with gpt-oss:20b. It's the cheapest in credits and handles most questions.
- Switch to gpt-oss:120b or nemotron-3-super when an answer needs more reasoning.
- Use nemotron-3-nano when you paste or reference very large files.
- Try beam for multi-step coding and agent work while it's free. It has a 1M context window and is built for software engineering, but check its credit usage before relying on it.
- Avoid nemotron-3-ultra on the free tier unless you need it, since it burns credits fastest.
Usage levels are Ollama's classification of how much of your free allowance each request consumes, so a "low" model lasts longer before you hit the limit. I couldn't find a usage level for beam, so check its model page.
Reference AI Chat Sessions¶
Ollama free tier cloud models API access
Ollama free tier cloud models API access - Different Account - Continuation of the previous session