Compare / head-to-head
DeepSeek V3.1-TerminusvsDeepSeek V3.2-Exp
DeepSeek V3.2-Exp leads 6 of 12 shared benchmarks. Both offer a 128K-token context window.
Benchmarks from cited public sources; pricing from official pages; status from official provider feeds.
Shared benchmarks
5 – 6
DeepSeek V3.2-Exp leads · 1 tied
Cheaper per token
—
list price, input + output
Larger context
Tie
both 128K tokens
Providers
DeepSeek
same provider
DeepSeek V3.1-Terminus
DeepSeek · released 2025-09-22
text
- Context
- 128K
- Max out
- —
- Input /1M
- —
- Output /1M
- —
- Cached /1M
- —
- Scores
- 12 · 5 core
DeepSeek V3.2-Exp
DeepSeek · released 2025-09-29
text
- Context
- 128K
- Max out
- —
- Input /1M
- —
- Output /1M
- —
- Cached /1M
- —
- Scores
- 17 · 7 core
Quality
Benchmark matrix
12 shared · 0 only DeepSeek V3.1-Terminus · 5 only DeepSeek V3.2-Exp| Benchmark | DeepSeek V3.1-Terminus | DeepSeek V3.2-Exp | Δ | Edge |
|---|---|---|---|---|
| Reported by both · 12 | ||||
| MMLU-Pro | 85% ↗ | 85% ↗ | 0 | Tie |
| Humanity's Last Exam (no tools) | 21.7% ↗ | 19.8% ↗ | +1.9 pt | DeepSeek V3.1-Terminus |
| LMArena Elo | 1417.2 ↗ | 1424.9 ↗ | -7.7 | DeepSeek V3.2-Exp |
| Aider-Polyglot | 76.1% ↗ | 74.5% ↗ | +1.6 pt | DeepSeek V3.1-Terminus |
| BrowseComp | 38.5% ↗ | 40.1% ↗ | -1.6 pt | DeepSeek V3.2-Exp |
| BrowseComp-zh | 45% ↗ | 47.9% ↗ | -2.9 pt | DeepSeek V3.2-Exp |
| GPQA-Diamond | 80.7% ↗ | 79.9% ↗ | +0.8 pt | DeepSeek V3.1-Terminus |
| LiveCodeBench | 74.9% ↗ | 74.1% ↗ | +0.8 pt | DeepSeek V3.1-Terminus |
| SimpleQA (agentic tool use) | 96.8% ↗ | 97.1% ↗ | -0.3 pt | DeepSeek V3.2-Exp |
| SWE Verified | 68.4% ↗ | 67.8% ↗ | +0.6 pt | DeepSeek V3.1-Terminus |
| SWE-bench Multilingual | 57.8% ↗ | 57.9% ↗ | -0.1 pt | DeepSeek V3.2-Exp |
| Terminal-bench | 36.7% ↗ | 37.7% ↗ | -1 pt | DeepSeek V3.2-Exp |
| Only DeepSeek V3.1-Terminus reports · 0 | ||||
| DeepSeek V3.1-Terminus reports nothing DeepSeek V3.2-Exp does not. | ||||
| Only DeepSeek V3.2-Exp reports · 5 | ||||
| AIME 2025 | not reported | 89.3% ↗ | — | — |
| HMMT 2025 | not reported | 83.6% ↗ | — | — |
| LMArena WebDev (LMArena) | not reported | 1271.8 ↗ | — | — |
| Vending-Bench 2 (Andon Labs) | not reported | 1034 ↗ | — | — |
| WeirdML (Håvard Tveit Ihle) | not reported | 39.5% ↗ | — | — |
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 DeepSeek V3.2-Exp 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 | DeepSeek V3.2-Exp | Edge |
|---|---|---|---|
| Context window | 128K tokens | 128K tokens | Tie |
| Max output | — | — | — |
| Input price / 1M | — | — | — |
| Output price / 1M | — | — | — |
| Cached input / 1M | — | — | — |
| Input + output / 1M Lower is cheaper. List prices; batch, tool and regional fees excluded. | — | — | — |
| Modalities | text | text | Tie |
| Released | 2025-09-22 | 2025-09-29 | — |
| Cited benchmark scores | 12 | 17 | — |
Reliability
DeepSeek status
Live from /statusMore matchups
DeepSeek V3.1-Terminus vs …
Models sharing the most benchmarksMore matchups
DeepSeek V3.2-Exp 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 DeepSeek V3.2-Exp on your own prompts with Respan evals, or route to either through one gateway key with automatic failover.