I have signed up to opencode go to try out their offering.
Currently the models they offer are:
I created a prompt to build an html page with an interactive chart to convey the affect of costs on investment returns over the long term. Full prompt can be found here.
The process was:
- In build mode,
Implement @prompt.md
Using OpenCode 1.2.27
Created a new opencode session for each model. Variant was kept as default.
Results
Here is a screenshot of the output of different models, they all look quite similar:
Results ordered by slowest to fastest.
| Model | Time | Tokens | Used | Cost |
|---|---|---|---|---|
| qwen3.7-plus | 3m 15s | 15,461 | 2% | $0.01 |
| glm-5.2 | 3m 6s | 27,904 | 3% | $0.14 |
| minimax-m3 | 1m 46s | 17,742 | 3% | $0.00 |
| deepseek-v4-pro | 1m 20s | 18,187 | 2% | $0.05 |
| mimo-v2.5-pro | 1m 17s | 16,023 | 2% | $0.04 |
| qwen3.7-max | 58.4s | 14,976 | 1% | $0.07 |
| mimo-v2.5 | 48.6s | 16,925 | 2% | $0.00 |
| deepseek-v4-flash | 43.6s | 23,048 | 2% | $0.00 |
| kimi-k2.7-code | 40.1s | 15,168 | 6% | $0.03 |
minimax-m3 failed the task – the infographic was incorrect.
mimo models failed to name the output file correctly
Conclusion
They all look quite similar. GLM-5.2 looks the best and has the most info – it was quite slow though.
Kimi-2.7-code was fast but not layed out well.
The Mimo and minimax models failed the task.
Deepseek-v4-flash is still one of the best for time and quality.
If you are interested in trying something like this yourself – check out opencode go (that is my referral link).