Qwen model lands in the OpenCode Go coding agent
- Source
- x.com
- Date
Live in OpenCode Go. Try it now! 👀 https://t.co/e7W2cHvABw
Context
Qwen, Alibaba's model team, says its new flagship Qwen4-Max is live in OpenCode Go, repeating OpenCode's own announcement. Go is a subscription sold alongside the OpenCode coding rather than the agent itself: a fixed monthly fee buys a capped amount of use across a curated list of models, reached with one API key. The Go documentation lists three plans, Go at $10 a month, Go Plus at $40 and Go Teams at $20 per member. It works in OpenCode and in other coding agents that identify themselves and send a session ID with each conversation, and Claude Code, Codex, Hermes and Pi are among the clients OpenCode says it has validated.
For Qwen4-Max, the subscription includes a small allowance charged at Qwen's own prices. Go counts $1.50 per million input and $4.50 per million output tokens, the list prices Qwen set at launch, plus $0.075 per million cached input tokens and $1.875 per million to write to the cache. Cached input is text the service has already processed and can reuse, such as the unchanged part of a conversation that an agent resends on every turn. Each model has its own monthly dollar limit, and for Qwen4-Max it is $15 on Go, $60 on Go Plus and $20 per member on Go Teams, with no more than 20% usable in any five hours and 50% in a week. OpenCode turns this into estimated requests by assuming each one carries 420 new input tokens, 66,000 cached tokens and 200 output tokens: about 460 requests per five hours and 2,310 a month on Go, and 1,840 and 9,240 on Go Plus. Those are OpenCode's illustrative figures, and real sessions will use more or fewer tokens.
Go also offers the previous Qwen3.8 Max under the same $15 monthly limit and the same token assumptions, and OpenCode estimates 810 requests a month for it against 2,310 for Qwen4-Max. Almost all of each assumed request is cached input, so the drop in the cached-read price from $0.25 to $0.075 per million accounts for most of that gap. The $15 limit sits at the low end of Go's range, where many models get $60, and the docs explain why some models get less: when a model is new, or its public pricing is already discounted, OpenCode has not negotiated a bulk discount or cheaper hosting, so subscribers get only a little more than they would by paying the provider directly. The docs do not name the models in that group, but a new model billed at its list price fits the description. Qwen4-Max in Go is a way to use Qwen's new model inside an existing coding setup with a modest included allowance, and heavy users will reach its limit well before those of the cheaper models. At that point Go blocks further requests to it, though the free models stay available, unless Extra Usage is turned on to bill the extra requests to a pay-as-you-go credit balance. Go Teams starts with Extra Usage on.
Checking sign-in…
Loading comments…




