Compare / head-to-head
DeepSeek V3.2vsLlama 3.3 70B Instruct
DeepSeek V3.2 leads 3 of 3 shared benchmarks. Both offer a 131K-token context window.
Benchmarks from cited public sources; pricing from official pages; status from official provider feeds.
Shared benchmarks
3 – 0
DeepSeek V3.2 leads
Cheaper per token
—
list price, input + output
Larger context
Tie
both 131K tokens
Provider uptime (30d)
100% · —
DeepSeek · Meta
DeepSeek V3.2
DeepSeek · released 2025-12-01
text
- Context
- 131K
- Max out
- 66K
- Input /1M
- —
- Output /1M
- —
- Cached /1M
- —
- Scores
- 8 · 7 core
Llama 3.3 70B Instruct
Meta · released 2024-12-06
text
- Context
- 131K
- Max out
- —
- Input /1M
- —
- Output /1M
- —
- Cached /1M
- —
- Scores
- 5 · 4 core
Quality
Benchmark matrix
3 shared · 4 only DeepSeek V3.2 · 1 only Llama 3.3 70B Instruct| Benchmark | DeepSeek V3.2 | Llama 3.3 70B Instruct | Δ | Edge |
|---|---|---|---|---|
| Reported by both · 3 | ||||
| GPQA Diamond | 82.4% ↗ | 50.5% ↗ | +31.9 pt | DeepSeek V3.2 |
| MMLU-Pro | 85% ↗ | 68.9% ↗ | +16.1 pt | DeepSeek V3.2 |
| LMArena Elo | 1424.8 ↗ | 1274 ↗ | +150.8 | DeepSeek V3.2 |
| Only DeepSeek V3.2 reports · 4 | ||||
| SWE-bench Verified | 73.1% ↗ | not reported | — | — |
| AIME 2025 | 93.1% ↗ | not reported | — | — |
| AIME 2026 | 94.2% ↗matharena | not reported | — | — |
| Humanity's Last Exam (no tools) | 25.1% ↗ | not reported | — | — |
| Only Llama 3.3 70B Instruct reports · 1 | ||||
| OTIS Mock AIME 2024-2025 (Epoch AI run) | not reported | 5.1% ↗epoch | — | — |
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 Llama 3.3 70B Instruct 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 | Llama 3.3 70B Instruct | Edge |
|---|---|---|---|
| Context window | 131K tokens | 131K tokens | Tie |
| Max output | 66K tokens | — | — |
| 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 | 2024-12-06 | — |
| Cited benchmark scores | 8 | 5 | — |
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
Llama 3.3 70B Instruct 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 Llama 3.3 70B Instruct on your own prompts with Respan evals, or route to either through one gateway key with automatic failover.