OpenCode Zen (free models)
Anomaly
The vendor marks selected models as Free, callable at no cost through an OpenAI-compatible API endpoint, no credit card required; the free scope is limited to the subset of model IDs carrying the -free suffix, distinct from the paid production versions of the same names
The operating entity is Anomaly (github.com/anomalyco) — OpenCode is its open-source coding-agent product line, and Zen is the model gateway that ships alongside it, not a third-party reseller. The free-model list rotates (mostly limited-time promotions during a new model's launch window, used to gather feedback); as of 2026-09-18, official documentation was re-checked: the previous Hy3 Free has been removed from the list, and Union Alpha Free was added (union-alpha, a stealth model with a zero-retention policy from the provider, joining Big Pickle as one of the free models without the -free suffix), and Muse Spark's version number was bumped from 1.2 to 1.3; the current full list is Big Pickle (big-pickle), Union Alpha Free (union-alpha), MiMo-V2.5 Free (mimo-v2.5-free), Ling 3.0 Flash Fin Free (ling-3.0-flash-fin-free), Nemotron 3 Ultra Free (nemotron-3-ultra-free), Nemotron 3.5 Lightning Free (nemotron-3.5-lightning-free), and Muse Spark 1.3 Contributor Free (muse-spark-1.3-contributor-free) — 7 in total; there's no guarantee any given model stays free long-term. Paid (non -free) models are billed pay-as-you-go: when the balance drops below $5, the card on file is auto-charged $20, with card fees passed through at cost (4.4% + $0.30/transaction); the vendor states there's no markup on model pricing itself.
The free models aren't limited to use within the opencode CLI: the base URL is https://opencode.ai/zen/v1, and any standard OpenAI-compatible client (e.g. @ai-sdk/openai-compatible) can call it directly. Registration requires generating an API Key at opencode.ai/auth. The official terms state clearly that conversation data from Big Pickle / MiMo-V2.5 / Hy3 and similar models may be used during their free period to improve the models; the NVIDIA-family terms (Nemotron 3 Ultra / 3.5 Lightning) go further, noting the data may be used to train NVIDIA's future products — if you're privacy-conscious, avoid submitting sensitive content to the free models.