OpenRouter
A global model aggregator: one API Key for many models. Part of the catalog is a free pool (IDs often end with :free, or marked Free), good for personal trials, comparing open-source models, and light roleplay. Official collection: Free Models; third-party summary: openairouter.net/free-models (unofficial — trust the OpenRouter console).
Free lists and RPM / daily caps change. Trust openrouter.ai/models and your account page.
Sites and entry points
| Resource | URL |
|---|---|
| Site | openrouter.ai |
| API Key | openrouter.ai/keys |
| Free Models collection | openrouter.ai/collections/free-models |
| Model plaza (filter Free) | openrouter.ai/models?q=free |
| Community summary (third-party) | openairouter.net/free-models |
| API Base URL | https://openrouter.ai/api/v1 (usually auto when MiniTavern picks OpenRouter) |
How to get free access
- Open openrouter.ai/keys, register, and create a Key (usually no credit card).
- Copy a Model ID from Free Models or the plaza (watch the
:freesuffix). - In MiniTavern pick platform OpenRouter, paste the Key, and enter that ID.
| Item | Details |
|---|---|
| Free path | IDs with :free, or $0 models in the collection; or auto-route openrouter/free |
| Rate limits (reference) | Often ~20 RPM / ~50 requests/day; a one-time top-up of about $10+ can raise free daily caps to ~1000/day (follow official rules) |
| Card required | Free tier usually no; top-up only to raise caps or call paid models |
| Privacy | Some free upstreams may use prompts to improve models; check OpenRouter privacy settings. Poolside Laguna free docs say data may be used for training |
Model list
Common text / vision IDs from the collection and public summaries (~2026-08); trust the live site. IDs are case-sensitive; do not drop :free unless that entry has no suffix.
Text · auto-route
| Model ID | Notes |
|---|---|
openrouter/free | Auto-picks an available free model per request; fastest start, quality varies |
Text · general chat
| Model ID | Notes |
|---|---|
meta-llama/llama-3.3-70b-instruct:free | General, multilingual |
meta-llama/llama-3.2-3b-instruct:free | Lightweight |
openai/gpt-oss-20b:free / openai/gpt-oss-120b:free | OpenAI open-weight MoE |
google/gemma-3-27b-it:free / google/gemma-3-12b-it:free / google/gemma-3-4b-it:free | Gemma 3 series |
google/gemma-4-31b-it:free / google/gemma-4-26b-a4b-it:free | Gemma 4 (some support vision) |
google/gemma-3n-e4b-it:free / google/gemma-3n-e2b-it:free | On-device–oriented light models |
qwen/qwen3-next-80b-a3b-instruct:free | Qwen3 chat |
z-ai/glm-4.5-air:free | GLM light, Chinese-friendly |
minimax/minimax-m2.5:free | MiniMax |
nousresearch/hermes-3-llama-3.1-405b:free | Hermes 3 |
arcee-ai/trinity-large-preview:free | Arcee Trinity Large Preview |
openrouter/elephant-alpha | OpenRouter-hosted text (if Free in collection/plaza) |
cognitivecomputations/dolphin-mistral-24b-venice-edition:free | Venice Uncensored variant (content policy is on you) |
Text · reasoning / Agent / coding
| Model ID | Notes |
|---|---|
qwen/qwen3-coder:free | Qwen3 Coder, code-oriented |
nvidia/nemotron-3-ultra:free (or current collection ID) | Nemotron 3 Ultra, long-context Agent |
nvidia/nemotron-3.5-lightning:free | High-throughput Agent |
nvidia/nemotron-3-super-120b-a12b:free | Nemotron 3 Super |
nvidia/nemotron-3-nano-30b-a3b:free / nvidia/nemotron-nano-9b-v2:free | Nano series |
poolside/laguna-s-2.1:free / poolside/laguna-xs-2.1:free (plaza IDs) | Poolside coding Agent; free tier may train on data |
cohere/north-mini-code:free (plaza) | Cohere North coding |
liquid/lfm-2.5-1.2b-instruct:free / liquid/lfm-2.5-1.2b-thinking:free | Liquid light |
dots-studio/dots3-note-preview:free (plaza) | Dots3-Note Preview |
Collection sort by usage changes; coding models can also be filtered Free under Coding.
Text · vision (multimodal input)
If the chat main model supports vision, try with App vision / multimodal:
| Model ID | Notes |
|---|---|
google/gemma-3-27b-it:free and other Gemma 3/4 | Text + image input |
nvidia/nemotron-nano-12b-v2-vl:free | Vision / documents |
nvidia/nemotron-3-nano-omni:free (plaza) | Multimodal perception |
meta-llama/llama-guard-4-12b:free | Safety classifier — usually not a chat main model |
Images / video
- Image gen: see OpenRouter Image Models; free or not follows each model’s pricing. MiniTavern image link can pick OpenRouter.
- Video: see Video Models; the App has no built-in video entry.
- Third-party lists may include Lyria-style music models — unrelated to chat / image gen; do not use as dialogue models.
How to fill in MiniTavern
General steps: Model link.
Chat
Path: Settings → Models → Model link → New
- Platform: OpenRouter (preset Base URL)
- Paste API Key
- Model ID e.g.
meta-llama/llama-3.3-70b-instruct:free, oropenrouter/free - Test → on success, set as default
You can also use Custom (OpenAI) with Base URL https://openrouter.ai/api/v1; prefer the App preset.
Image gen
Path: Settings → Image gen → Image gen link → platform OpenRouter, pick a plaza ID that supports images and is free (or billing you accept). More reliable free image gen: Agnes AI, AIHubMix.
Troubleshooting
| Symptom | Fix |
|---|---|
HTTP 429 / Provider returned error | Free pool full or rate-limited: switch :free, wait, or BYOK |
| Access fails in mainland China | Official API may need a proxy; or use mainland-direct vendors / AIHubMix |
| Model 404 / unavailable | That :free retired: copy a fresh ID from the collection |
More: Error codes.
Related
- Free cloud overview
- FreeLLM directory — more free vendors
- AIHubMix — subsidized
*-free(including some closed-source mirrors) - Agnes AI · OpenCode Zen
- Model link
- Free image gen