OpenCode Go Bundles DeepSeek V4 Flash, Qwen3.8 Max, and GLM-5.2 for $10/Month

Just yesterday I shared my experience using DeepSeek V4 Flash with OMP/Pi, and this morning when I opened the DeepSeek backend, I saw a notification:

We plan to raise the overall pricing of the DeepSeek API service in the near future, and the increase is expected to be significant.

This message is now visible, and it can also be found in the official documentation:

What a coincidence.

Just yesterday, I was recommending in a tech group that friends who hadn't tried it should give it a go. DeepSeek V4 Flash's coding ability is quite good, it's fast, and the price is low. Connecting it to Coding Agents like OMP and Pi makes for a very smooth experience.

And right after I recommended it, the official announcement came that they're preparing to raise prices.

However, this is understandable. DeepSeek V4 Flash has indeed been phenomenally popular recently, and its current price is almost absurdly low. Personally, even if it doubles or triples from the current base, I'd probably still find it acceptable, given how low the starting price is.

If you're worried about the cost after the price hike, you can also check out OpenCode Go.

It's currently only $5 for the first month, and $10 per month after that. Besides DeepSeek V4 Flash, you can also use a batch of mainstream models like Qwen3.8 Max, Kimi K3, GLM-5.2, MiniMax M3, and Kimi K2.7 Code.

The value for money is indeed quite high.

Yesterday, a reader also asked in the comments what the difference is between OpenCode and Pi when both are connected to DeepSeek. I was planning to talk about this, but I got distracted playing games and forgot.

By the way, this is not an ad. Under yesterday's article introducing OMP and Pi, a reader asked, "Agai, if you're taking sponsorship money, just blink." It really isn't! Same goes for this OpenCode piece; I just thought this subscription plan was interesting after seeing it, so I casually compiled the relevant info and setup method.

Today, I'm making up for it.

What is OpenCode Go?

OpenCode is an open-source AI Coding Agent that can be used in the terminal, desktop app, and IDE. Its overall form is quite similar to Claude Code and Codex CLI; it can read code, modify files, execute commands, and invoke various development tools.

OpenCode itself is free, and you can also connect it to APIs from different providers.

OpenCode officially provides two model services:

• OpenCode Zen, billed based on actual Token usage;
• OpenCode Go, a monthly subscription that allows you to use multiple models with one subscription.

You can think of OpenCode Go as a model plan specifically prepared for Coding Agents. After subscribing, you get an API Key, and through this Key, you can use the models that OpenCode has already adapted.

The price is $5 for the first month, and $10 per month after that.

What models can you use for $10?

As of August 6, 2026, OpenCode Go supports quite a few models, as shown in the table below:

I saw a bill shared by a friend. After connecting DeepSeek V4 Flash to Codex via OpenCode Go, a single request carried 156,000 to 190,000 input tokens, with output ranging from a few hundred to over a thousand tokens. The final cost was only $0.0006 to $0.0014.

The reason it's so low is mainly because a large amount of context hits the cache. DeepSeek V4 Flash's cache read price is only $0.0028 per million tokens, which is especially suitable for Coding Agent scenarios where code and context are repeatedly carried.

Of course, 158,000 requests is just an estimate given by OpenCode based on typical usage and does not represent unlimited calls. OpenCode Go currently also has limits of $12 per 5 hours, $30 per week, and $60 per month. The actual number of requests you can make varies depending on the model.

Qwen3.8 Max is a very typical example. Although it is also included in the $10 plan, its converted monthly quota is only $15, making it more suitable for handling complex planning and difficult problems. It is not recommended to use it as the default model all the time.

For the complete model list and the latest quotas, you can check the OpenCode Go official documentation.

How to use OpenCode Go?

Installation, subscription, and connection can all be done in a few minutes.

First, install OpenCode. For macOS and Linux, you can execute:

curl -fsSL https://opencode.ai/install | bash

If you already have Node.js installed, you can also install it via npm:

npm install -g opencode-ai

After installation, enter your project directory and start OpenCode:

cd /path/to/project
opencode

Then, in OpenCode, enter:

/connect

In the Provider list, select OpenCode Go. OpenCode will prompt you to go to the console to log in, subscribe to Go, and copy the API Key. Just paste the generated Key back in.

After a successful connection, enter:

/models

At this point, you can see all the models supported by the Go plan, including:

opencode-go/deepseek-v4-flash
opencode-go/qwen3.8-max
opencode-go/kimi-k3
opencode-go/glm-5.2
opencode-go/minimax-m3
opencode-go/kimi-k2.7-code

After selecting a model, you can start using it directly in the current project. If you want to switch models midway, just execute /models again; there's no need to reconfigure the API Key.

The OpenCode Go API Key can also be used with other Coding Agents that support custom Providers. Someone connected DeepSeek V4 Flash to Codex via opencodex, which is how the low-cost bill mentioned earlier was generated.

However, different models may use different API protocols. DeepSeek V4 Flash uses an OpenAI-compatible interface, while Qwen3.8 Max uses the Anthropic Messages interface. If you just want to get started as quickly as possible, selecting the model directly through OpenCode is the easiest way.

Summary

OpenCode Go is a monthly subscription plan provided by OpenCode. One API Key can be used to access multiple models like DeepSeek V4 Flash, Qwen3.8 Max, and Kimi K3. The setup process is also not complicated: install OpenCode, enter the API Key via /connect, and then use /models to select or switch models.

This article mainly compiled the common models in the current plan, their corresponding quotas, and the complete workflow from installation to model selection. The model list and quotas may be adjusted; please refer to the official OpenCode Go documentation for the actual information.