In the author's one comparison, Terra cost less and took less time, while Sol delivered an interactive cityscape whose day and night views could be switched with a time-of-day slider. This is a single result from one post and cannot establish the models' overall capabilities.
Tasks this comparison can inform: Comparing how day and night, boats, cars, and other elements are rendered under the same New York city SVG prompt, as well as cost, time, and interactive completeness.
Claims it cannot support: Generalizing to other coding tasks, complex projects, repeat-run reliability, overall value, or general model capabilities.
Model versions covered: GPT-6.1 Terra and GPT-6.1 Sol (as labeled in the post).
Test environment or client: Not specified.
Reasoning level and parameters: Not specified.
The author says both models used the same prompt to generate an animated New York SVG featuring day and night, boats, and cars, and reports each model's price, elapsed time, token count, and output characteristics. The post does not disclose the full prompt, client or harness, runtime environment, timing or pricing methodology, or reasoning level. This is one run reported by the author and has not been independently reproduced.
| Model | Cost reported by author | Time | Tokens | Output as described by author |
|---|---|---|---|---|
| GPT-6.1 Terra | $0.12 | 4 min 32 sec | 25K | A polished, postcard-style static illustration; the post reports no corresponding interaction and says the day/night change was less pronounced |
| GPT-6.1 Sol | $0.47 | 8 min 20 sec | 49K | An interactive cityscape with a working time-of-day slider; the post says it shows the day/night transition more clearly |
Based on the figures in the post, Sol cost about 3.9 times as much as Terra, took about 1.8 times as long, and used about 2 times as many tokens. These ratios are calculated from the reported values for one run and do not represent a stable difference.
Shared task reported by the author: an animated New York SVG with day/night, boats, and cars.
Terra: $0.12, 4 min 32 sec, 25K tokens; the output was described as a polished, postcard-style illustration.
Sol: $0.47, 8 min 20 sec, 49K tokens; the output had a working day/night time-of-day slider.
Author's assessment: Terra was cheaper and faster; Sol showed the day/night change more clearly.
All descriptions of output quality and functionality are relayed from the post. They were not inferred from the attached video, and the finished outputs were not independently inspected.
This record supports a limited conclusion: in the author's reported SVG task, Terra used less money, time, and fewer tokens, while Sol implemented a clear interactive control. The post does not provide enough detail to verify the prompt, execution environment, or measurement methods, so the difference cannot be attributed to the models themselves or generalized to cost, speed, or completion on other tasks.
There is not enough information for strict reproduction: the original post does not provide the full prompt, test harness, run parameters, or measurement methodology. To verify the result, those details would need to be supplied and both models run repeatedly with the same environment and prompt. This record is not an independent reproduction experiment.
GPT-6.1 Terra
