OpenCode Free Models - 4 Free AI Models in the Client (2026)
OpenCode offers 4 limited-time free models covering DeepSeek, GLM, MiniMax, and LongCat, all at $0/M tokens. They are suitable for coding, agent, and long-context tasks inside the OpenCode client.
https://opencode.ai/inference/openai/v1/chat/completions directly returns 403 FreeTierError. The free tier also has an undisclosed daily request limit, and LongCat 2.5 has a separate 100 requests/day limit.Model List (4)
| Model | Context | Max output | Input | Data terms |
|---|---|---|---|---|
| DeepSeek V4.1 Flash Free | 200K | 32K | Text + image | Zero retention, no training |
| GLM-5.3-Flash Free | 200K | 32K | Text + image + video | Zero retention, no training |
| MiniMax M3.1 Flash Free | 1M | 512K | Text + image + video | Zero retention, no training |
| LongCat 2.5 Free | 1M | 262K | Text + image | Zero retention, China-hosted, 100 RPD |
Why only four? The OpenCode API currently lists 12 free chat models, but Nemotron 3.5 Super VL and Nemotron 3.5 Lightning are already listed on OpenRouter here. Inkling 2 Small and North Mini Code are still free inside OpenCode on a limited-time basis, but their OpenRouter free versions have ended, so we do not list them separately yet. Flint 40B overlaps heavily with Recursive Flint; Big Pickle and Iceman are stealth models; Beam is not available in mainland China; and MiMo, Nemotron, and others have weaker data terms. We keep only the four zero-retention, complementary, and relatively stable models.
Free Rules
- Limited-time free: all models are marked “free on OpenCode for a limited time” and may become paid or be removed at any time.
- Daily request limit: the free tier has an undisclosed daily request limit shared by all free models; LongCat 2.5 has an additional 100 requests/day limit.
- Client only: free models work only in OpenCode CLI / Desktop / IDE; direct API calls return 403.
- Data terms: all four listed models have zero retention and are not used for training; other free models may log data or be used to improve models.
- Regional restrictions: Beam Free is not available in mainland China (including Hong Kong and Macau); the four listed models have no similar restriction.
How to Use
- Install OpenCode:
curl -fsSL https://opencode.ai/v2/install | bash(orbrew install anomalyco/tap/opencode-v2). - Run
/modelsin OpenCode and select a model such asopencode/deepseek-v4.1-flash-free. - Or pin the model in your project’s
opencode.json:
{
"$schema": "https://opencode.ai/config.json",
"model": "opencode/deepseek-v4.1-flash-free"
}opencode run --model opencode/minimax-m3.1-flash-free "Explain this code"