GPT-6.1 Terra vs Hy4
aka OpenAI GPT-6.1 Terra vs Tencent Hy4
- CheaperEven1–1
- FasterEven1–1
As stated in each post, not a controlled benchmark.
Posts
4
3D football game “4th & Goal”, four models: Hy4, GPT-6.1 Terra, GPT-6.1 Sol, DeepSeek V4.1 Flash 1017. Posted by @sshayahmet.
Explorable Romanesque cathedral, two models: GPT-6.1 Terra, Hy4. Posted by @doe_naps.
Medieval castle in one HTML file, two models: Hy4, GPT-6.1 Terra. Posted by @marcthecreatorr.
Same prompt, separate runs
2Not a controlled comparison: prompts, tools and time differ.
About this matchup
GPT-6.1 Terra vs Hy4 in real side-by-side tests: 4 posts from X where both got the same prompt; 3 posts report cost or time. Judge coding ability from the actual outputs.
GPT-6.1 Terra vs Hy4: which is better for coding?
Which is better for coding, GPT-6.1 Terra or Hy4?
None of the 4 posts here names one of the two as the better result, so judge the outputs above for the work you care about.
Which is cheaper, GPT-6.1 Terra or Hy4?
2 posts state a cost for both: each was cheaper in 1. These are the costs the posts state for each run, not list prices.
Which is faster, GPT-6.1 Terra or Hy4?
2 posts state a time for both: each was faster in 1. Times are as stated in each post; tools and settings differ between posts.
Summary
GPT-6.1 Terra and Hy4 appear together in 4 posts on X: cheetah scene in OpenCode, three models (@LatentOperatorX), 3D football game “4th & Goal”, four models (@sshayahmet), explorable Romanesque cathedral, two models (@doe_naps) and medieval castle in one HTML file, two models (@marcthecreatorr). 2 of them are part of larger lineups with other models. In @LatentOperatorX's cheetah scene in OpenCode, three models, GPT-6.1 Terra was cheaper ($0.04 vs $0.08) and faster (2.1 min vs 3.9 min), as reported in the post. In @marcthecreatorr's medieval castle in one HTML file, two models, Hy4 was cheaper ($0.03 vs $0.44) and faster (6 min vs 16 min), as reported in the post.
What each post reports
| Post | Model | Time | Cost | Tokens | Author’s verdict | Post |
|---|---|---|---|---|---|---|
| Cheetah scene in OpenCode, three models@LatentOperatorX | Hy4 | 3.9 min | $0.08 | — | DeepSeek is the poster’s clear visual winner; the poster is disappointed by Hy4. | X post |
| Cheetah scene in OpenCode, three models@LatentOperatorX | GPT-6.1 Terra | 2.1 min | $0.04 | — | — | X post |
| 3D football game “4th & Goal”, four models@sshayahmet | Hy4 | 1 h 36 min | — | — | — | X post |
| 3D football game “4th & Goal”, four models@sshayahmet | GPT-6.1 Terra | — | — | — | — | X post |
| Explorable Romanesque cathedral, two models@doe_naps | GPT-6.1 Terra | — | — | — | — | X post |
| Explorable Romanesque cathedral, two models@doe_naps | Hy4 | — | — | — | Hy4 did the exterior well but didn’t quite get the interior. | X post |
| Medieval castle in one HTML file, two models@marcthecreatorr | Hy4 | 6 min | $0.03 | — | Hy4 isn’t at Claude Sonnet 5.5 or DeepSeek V4.1 Flash 1017 level in 3D coding yet, but it is very fast and cheap. | X post |
| Medieval castle in one HTML file, two models@marcthecreatorr | GPT-6.1 Terra | 16 min | $0.44 | — | — | X post |
Only numbers stated in each post. “—” means the post doesn’t say.
Quick facts
| Model | Vendor | Released | Input $/1M tokens | Output $/1M tokens | Context (tokens) | Official links |
|---|---|---|---|---|---|---|
| GPT-6.1 Terra | OpenAI | 2026-10-26 | Announcement | |||
| Hy4 | Tencent | 2026-11-04 | Announcement |
From the vendors’ official pages, checked 2026-11-04. Standard API list prices; each value links its source and a blank cell could not be verified. The archive’s own numbers above are what each post states.
Who won where
3- Cheetah scene in OpenCode, three modelsDeepSeek is the poster’s clear visual winner; the poster is disappointed by Hy4.— the author’s view, @LatentOperatorX (paraphrased)
- Explorable Romanesque cathedral, two modelsHy4 did the exterior well but didn’t quite get the interior.— the author’s view, @doe_naps (paraphrased)
- Medieval castle in one HTML file, two modelsHy4 isn’t at Claude Sonnet 5.5 or DeepSeek V4.1 Flash 1017 level in 3D coding yet, but it is very fast and cheap.— the author’s view, @marcthecreatorr (paraphrased)
Models
Related matchups
Claude Opus 5.5 vs GPT-6.1 Terra96 posts · 209.6K viewsGPT-6.1 Sol vs GPT-6.1 Terra91 posts · 238.3K viewsDeepSeek V4.1 Flash 1017 vs GPT-6.1 Terra54 posts · 140.5K viewsGPT-6 Astra vs GPT-6.1 Terra48 posts · 85.4K viewsClaude Sonnet 5.5 vs GPT-6.1 Terra32 posts · 53.5K viewsGemini 4 Argon vs GPT-6.1 Terra20 posts · 24.3K views