OpenCode adds an anonymous million-context model for free, but it is unavailable in the EU

A new unattributed option has appeared in OpenCode's model list: Pulse Alpha Free. The official docs label it a stealth model, list a 1 million token context, and say it is temporarily free to gather feedback and improve the model. The selector path is opencode/pulse-alpha-free. There is no verified attribution and no independent benchmark. What can be confirmed now is that the listing really exists; which lab made it remains a guess.

The screenshot shows that Pulse Alpha Free is not the only free model. GLM-5.3-Flash Free, DeepSeek V4.1 Flash Free, MiMo-V2.6-Flash Free, Nemotron 3 Ultra Free, Big Picle, Pulse Alpha Free, Beam Free, Flint 40B Free, and Jev 1.13 Free are all on the page. The concentration of free models means low-cost experimentation for developers and feedback and data collection for the platform. At the bottom of the page there is also a note that free models can be hidden.

There is one limitation to look at first: the docs mention that conversations may be retained for model improvement. Another limitation is regional access; the EU and several other regions are blocked. Before testing, confirm two things: the data terms and your network location. A million-token context sounds large and can fit long codebases or long documents, but a long context does not equal good usability. Tool calling, retrieval quality, latency, and stability all need real testing. When there are no independent benchmarks, don't treat 'one million' as a conclusion about capability.

One commenter replied that free is great, but the real test is whether it can survive the 'free trial graveyard.' That line fits right now. With an anonymous model opened for free, users get a taste, and the platform gets feedback and possibly retained conversations. Whether it is worth migrating workflows depends on results, not labels.