whatmodel.sh / Hunyuan-TurboS (20250416)
Hunyuan-TurboS (20250416)
↗ share
+ contribute
Hunyuan-TurboS (20250416) is a closed-weights language model from Tencent, released April 2025. Its published size is 560B parameters (56B active per token).
Catalogue identity notes
Comparison-only. Arena entry 'hunyuan-turbos-20250416' (text). LMArena's leaderboard table (arena-leaderboard space @6f2eeeb) labels 'hunyuan-turbos-20250416' as 'Hunyuan-Turbos-20250416' (Tencent, Proprietary), linking https://cloud.tencent.com/document/product/1729/104753. Released 2025-04: Tencent Cloud Hunyuan version changelog (document 1729/97765) / date in the API id. Exact dated Tencent Cloud API snapshot; Tencent's TurboS technical report (arXiv 2505.15431) names this snapshot as the model it ran on Chatbot Arena and gives 560B total / 56B active (hybrid Transformer-Mamba MoE). Retired.
$ how it compares
The LLMs ranked either side of Hunyuan-TurboS (20250416), out of 1132 with a score, then more from Tencent.
| # | model | score | where | ||
|---|---|---|---|---|---|
| 224 |  Gemini 2.5 Flash Preview (09-2025) | 63±1 | comparison only | ||
| 225 |  Qwen3-Next 80B-A3B Instruct | 63±1 | api $0.34/M +local | ||
| 226 | Hunyuan-TurboS (20250416) | 63±3 | comparison only | ||
| 227 |  Granite 4.2 30B | 63±1 | api $0.28/M +local | ||
| 228 | Hunyuan-Vision-1.5-Thinking | 63±3 | api price unknown | ||
| more from Tencent | |||||
| 16 |  Hy4 preview (770B-A49B) | 79±1 | api $1.25/M +local | ||
| 57 | Hunyuan Hy3 (295B-A21B) | 74±1 | api $0.23/M +local | ||
| 139 | Hy3 preview (295B-A21B) | 68±1 | local | ||
$ what it's good at
chat & writing▁▁▁▃▅▅▆█▆▆▆▂ #168 of 579 * · 95% of its score · 1 test
Beats LongCat 2.0 (1.6T-A48B), o3-mini, K-EXAONE (236B-A23B) here, though they rank higher overall. Trails LFM2.5 8B-A1B, LFM2.5 2.6B, Nemotron-Cascade-8B here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| LMArena (text) | 1382.7 ✎ | ▁▁▁▂▃▃▃▆▆▇█▃ beats 221/379 | 95% |
$ where to run it
No verified API or download recorded yet · add one
$ how the score was worked out · LMArena leaderboard dataset 95%
Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works
| source | results | share of the score |
|---|---|---|
 LMArena leaderboard dataset | 1 | 95% |
| prior · pull toward the middle | 5% | |
| weighted average | = 63.1 |
$ improve this page
Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.
data updated · sources and licences · json