Compare / head-to-head
DeepSeek V3.1-Terminusvs
Hy3
Hy3 leads 4 of 4 shared benchmarks. Hy3 has the larger context window (256K tokens).
Benchmarks from cited public sources; pricing from official pages; status from official provider feeds.
Shared benchmarks
0 – 4
Hy3 leads
Cheaper per token
—
list price, input + output
Larger context
Hy3
256K tokens
Providers
2 providers
DeepSeek · Tencent Hunyuan
DeepSeek V3.1-Terminus
DeepSeek · released 2025-09-22
text
- Context
- 128K
- Max out
- —
- Input /1M
- —
- Output /1M
- —
- Cached /1M
- —
- Scores
- 12 · 5 core
Quality
Benchmark matrix
4 shared · 8 only DeepSeek V3.1-Terminus · 40 only Hy3| Benchmark | DeepSeek V3.1-Terminus | Hy3 | Δ | Edge |
|---|---|---|---|---|
| Reported by both · 4 | ||||
| Humanity's Last Exam (no tools) | 21.7% ↗ | 37% ↗ | -15.3 pt | Hy3 |
| LMArena Elo | 1417.2 ↗ | 1440.6 ↗ | -23.4 | Hy3 |
| BrowseComp | 38.5% ↗ | 84.2% ↗ | -45.7 pt | Hy3 |
| SWE-bench Multilingual | 57.8% ↗ | 75.8% ↗ | -18 pt | Hy3 |
| Only DeepSeek V3.1-Terminus reports · 8 | ||||
| MMLU-Pro | 85% ↗ | not reported | — | — |
| Aider-Polyglot | 76.1% ↗ | not reported | — | — |
| BrowseComp-zh | 45% ↗ | not reported | — | — |
| GPQA-Diamond | 80.7% ↗ | not reported | — | — |
| LiveCodeBench | 74.9% ↗ | not reported | — | — |
| SimpleQA (agentic tool use) | 96.8% ↗ | not reported | — | — |
| SWE Verified | 68.4% ↗ | not reported | — | — |
| Terminal-bench | 36.7% ↗ | not reported | — | — |
| Only Hy3 reports · 40 | ||||
| SWE-bench Verified | not reported | 78% ↗ | — | — |
| GPQA Diamond | not reported | 90.4% ↗ | — | — |
| AA-LCR | not reported | 73.4% ↗ | — | — |
| Apex-Agent | not reported | 25.6% ↗ | — | — |
| ArxivMath | not reported | 52.2% ↗ | — | — |
| CL-bench | not reported | 23.8% ↗ | — | — |
| CL-bench Life | not reported | 17% ↗ | — | — |
| ClawEval | not reported | 68.5% ↗ | — | — |
| CMT-Benchmark | not reported | 37.8% ↗ | — | — |
| DeepSearchQA | not reported | 91% ↗ | — | — |
| DeepSWE | not reported | 28% ↗ | — | — |
| e-bench | not reported | 50.2% ↗ | — | — |
| FrontierScience-Olympiad | not reported | 74.8% ↗ | — | — |
| FrontierScience-Research | not reported | 21.3% ↗ | — | — |
| HLE (with tools, text-only) | not reported | 53.2% ↗ | — | — |
| HorizonMath | not reported | 7.1% ↗ | — | — |
| Hy-Backend 2.0 | not reported | 25% ↗ | — | — |
| Hy-CompanyBench | not reported | 41.7% ↗ | — | — |
| Hy-Euler Pro | not reported | 24.2% ↗ | — | — |
| Hy-FinModelBench | not reported | 69% ↗ | — | — |
| Hy-Math | not reported | 60.9% ↗ | — | — |
| Hy-SkillsWorld | not reported | 45.8% ↗ | — | — |
| Hy-SWE Max | not reported | 49% ↗ | — | — |
| IMOAnswerBench | not reported | 90% ↗ | — | — |
| LMArena Agent (LMArena) | not reported | -0.0538 ↗ | — | — |
| LMArena WebDev (LMArena) | not reported | 1507.4 ↗ | — | — |
| MathArena Apex | not reported | 38.7% ↗ | — | — |
| MCP Atlas | not reported | 79.1% ↗ | — | — |
| NL2Repo | not reported | 45.6% ↗ | — | — |
| PHYBench | not reported | 77.4% ↗ | — | — |
| ProdBench | not reported | 23% ↗ | — | — |
| SkillsBench | not reported | 55.3% ↗ | — | — |
| SuperChem | not reported | 54.9% ↗ | — | — |
| SWE-bench Pro | not reported | 57.9% ↗ | — | — |
| Terminal-Bench 2.1 | not reported | 71.7% ↗ | — | — |
| Toolathlon | not reported | 48.5% ↗ | — | — |
| Toolathlon-Verified (HKUST) | not reported | 56.8% ↗ | — | — |
| USAMO 2026 | not reported | 72% ↗ | — | — |
| WideSearch | not reported | 76.4% ↗ | — | — |
| WildClawBench | not reported | 53.6% ↗ | — | — |
Scores tagged "epoch" or "matharena" are independent runs, used only where the lab has not published its own; ⚠ marks rows MathArena flags as released after the competition. Higher is better on every row. Δ is DeepSeek V3.1-Terminus minus Hy3 in the benchmark's own unit. "Not reported" means the lab has not published that figure; it is not a zero. ↗ opens the source.
Specs & pricing
Side by side
Official pricing pages and model cards| Spec | DeepSeek V3.1-Terminus | Hy3 | Edge |
|---|---|---|---|
| Context window | 128K tokens | 256K tokens | Hy3 |
| Max output | — | — | — |
| Input price / 1M | — | $0.132 ↗ | — |
| Output price / 1M | — | $0.528 ↗ | — |
| Cached input / 1M | — | $0.033 ↗ | — |
| Input + output / 1M Lower is cheaper. List prices; batch, tool and regional fees excluded. | — | $0.66 | — |
| Modalities | text | text | Tie |
| Released | 2025-09-22 | 2026-07-06 | — |
| Cited benchmark scores | 12 | 45 | — |
Reliability
Provider status
Live from /statusMore matchups
DeepSeek V3.1-Terminus vs …
Models sharing the most benchmarksMore matchups
Hy3 vs …
Models sharing the most benchmarksBuilt by Respan
Which one wins on your data?
Public benchmarks are a starting point. Run DeepSeek V3.1-Terminus and Hy3 on your own prompts with Respan evals, or route to either through one gateway key with automatic failover.