Compare / head-to-head
DeepSeek V3.2vsMAI-Thinking-1
MAI-Thinking-1 leads 4 of 5 shared benchmarks. MAI-Thinking-1 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
MAI-Thinking-1 leads · 1 tied
Cheaper per token
—
list price, input + output
Larger context
MAI-Thinking-1
256K tokens
Provider uptime (30d)
100% · —
DeepSeek · Microsoft
DeepSeek V3.2
DeepSeek · released 2025-12-01
text
- Context
- 131K
- Max out
- 66K
- Input /1M
- —
- Output /1M
- —
- Cached /1M
- —
- Scores
- 8 · 7 core
MAI-Thinking-1
Microsoft · released 2026-08-12
text
- Context
- 256K
- Max out
- 64K
- Input /1M
- —
- Output /1M
- —
- Cached /1M
- —
- Scores
- 15 · 7 core
Quality
Benchmark matrix
5 shared · 2 only DeepSeek V3.2 · 10 only MAI-Thinking-1| Benchmark | DeepSeek V3.2 | MAI-Thinking-1 | Δ | Edge |
|---|---|---|---|---|
| Reported by both · 5 | ||||
| SWE-bench Verified | 73.1% ↗ | 73.5% ↗ | -0.4 pt | MAI-Thinking-1 |
| GPQA Diamond | 82.4% ↗ | 84.2% ↗ | -1.8 pt | MAI-Thinking-1 |
| MMLU-Pro | 85% ↗ | 85% ↗ | 0 | Tie |
| AIME 2025 | 93.1% ↗ | 97% ↗ | -3.9 pt | MAI-Thinking-1 |
| AIME 2026 | 94.2% ↗matharena | 94.5% ↗ | -0.3 pt | MAI-Thinking-1 |
| Only DeepSeek V3.2 reports · 2 | ||||
| Humanity's Last Exam (no tools) | 25.1% ↗ | not reported | — | — |
| LMArena Elo | 1424.8 ↗ | not reported | — | — |
| Only MAI-Thinking-1 reports · 10 | ||||
| AdvancedIF | not reported | 85% ↗ | — | — |
| BFCL v3 | not reported | 72% ↗ | — | — |
| GraphWalks (<=128K) | not reported | 90% ↗ | — | — |
| HMMT February 2026 | not reported | 84.9% ↗ | — | — |
| IFBench | not reported | 69% ↗ | — | — |
| LiveCodeBench v6 | not reported | 87.7% ↗ | — | — |
| MultiChallenge | not reported | 53% ↗ | — | — |
| SimpleQA Verified | not reported | 31% ↗ | — | — |
| SWE-bench Pro | not reported | 52.8% ↗ | — | — |
| Terminal-Bench 2.0 | not reported | 46% ↗ | — | — |
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 V3.2 minus MAI-Thinking-1 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 V3.2 | MAI-Thinking-1 | Edge |
|---|---|---|---|
| Context window | 131K tokens | 256K tokens | MAI-Thinking-1 |
| Max output | 66K tokens | 64K tokens | DeepSeek V3.2 |
| 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 | 2025-12-01 | 2026-08-12 | — |
| Cited benchmark scores | 8 | 15 | — |
Reliability
Provider status
Last 60 days · refreshed daily on this page · live on /statusMore matchups
DeepSeek V3.2 vs …
Models sharing the most benchmarksMore matchups
MAI-Thinking-1 vs …
Models sharing the most benchmarksBuilt by Respan
Which one wins on your data?
Public benchmarks are a starting point. Run DeepSeek V3.2 and MAI-Thinking-1 on your own prompts with Respan evals, or route to either through one gateway key with automatic failover.