DDeepSeek · Released Dec 1, 2025
DeepSeek V3.2
DeepSeek-V3.2 is a large language model designed to harmonize high computational efficiency with strong reasoning and agentic tool-use performance. It introduces DeepSeek Sparse Attention (DSA), a fine-grained sparse attention mechanism...
- Top 16%Aider Polyglot
- Top 21%Fiction.LiveBench
- Released
- Dec 1 2025
- Context window
- 164K 66K out
- Input / 1M tokens
- $0.28 from Free
- Output / 1M tokens
- $0.42 from Free
Where to run it
43 providers compared
Free to try on iFlow — free endpoints are often limited-time and rate-limited. Included in Alibaba Token Plan, Alibaba Token Plan (China). Cheapest: GMICloud at $0.21 / $0.31 per 1M input / output tokens — 93% less than the priciest option.
| Provider | Input | Output | Cache read | Context | Max output | Uptime | vs. best |
|---|---|---|---|---|---|---|---|
| iFlowfree · may be limited-time | Free | Free | — | 128K | 64K | — | Free |
| Alibaba Token Plansubscription plan | Plan | Plan | — | 131K | 66K | — | Plan |
| Alibaba Token Plan (China)subscription plan | Plan | Plan | — | 131K | 66K | — | Plan |
| GMICloudFP8 | $0.21 | $0.31 | $0.022 | 164K | 147K | 99.7%24h | Cheapest |
| TokenGo | $0.22 | $0.33 | $0.06 | 128K | 64K | — | +4% |
| AtlasCloudFP8 | $0.26 | $0.38 | $0.13 | 164K | 147K | 99.9%24h | +24% |
| DeepInfraFP4 | $0.26 | $0.38 | $0.13 | 164K | 16K | 99.6%24h | +24% |
| Meganova | $0.26 | $0.38 | $0.13 | 164K | 164K | — | +24% |
| DevPass (LLM Gateway) | $0.26 | $0.38 | $0.13 | 164K | 16K | — | +24% |
| Poe | $0.26 | $0.38 | $0.13 | 128K | — | — | +25% |
| Venice | $0.27 | $0.39 | $0.13 | 160K | 33K | 99.5%24h | +27% |
| LLM Gateway | $0.27 | $0.40 | $0.14 | 164K | 66K | — | +29% |
| Abacus | $0.27 | $0.40 | — | 128K | 8K | — | +29% |
| Helicone | $0.27 | $0.41 | — | 164K | 66K | — | +30% |
| Alibaba (China) | $0.27 | $0.42 | — | 164K | 66K | — | +31% |
| Hugging Face | $0.28 | $0.40 | — | 164K | 66K | — | +32% |
| Merge Gateway | $0.28 | $0.40 | $0.13 | 128K | 33K | — | +32% |
| BaiduFP8 · no tool calling | $0.28 | $0.42 | $0.028 | 131K | 66K | 96.3%24h | +34% |
| Kilo Gateway | $0.28 | $0.42 | $0.028 | 131K | 66K | — | +34% |
| Vivgrid | $0.28 | $0.42 | $0.03 | 128K | 128K | — | +34% |
| 302.AI | $0.29 | $0.43 | — | 128K | 128K | — | +39% |
| Ofox | $0.29 | $0.43 | $0.06 | 128K | 32K | — | +39% |
| Qiniu | $0.29 | $0.43 | $0.029 | 128K | 32K | — | +39% |
| ZenMux | $0.29 | $0.44 | $0.029 | 128K | 8K | — | +41% |
| TensorX | $0.30 | $0.50 | $0.075 | 164K | 164K | — | +49% |
| Venice AI | $0.33 | $0.48 | $0.16 | 160K | 33K | — | +57% |
| SiliconFlowFP8 | $0.26 | $0.80 | $0.14 | 164K | 147K | 99.4%24h | +68% |
| Melious | $0.35 | $0.58 | $0.093 | 164K | 64K | — | +73% |
| DigitalOceanno tool calling | $0.30 | $0.96 | $0.09 | 164K | 147K | 99.8%24h | +99% |
| NanoGPTno tool calling | $0.50 | $1.00 | $0.25 | 164K | 66K | — | +167% |
| SiliconFlow (China) | $0.60 | $0.90 | $0.06 | 164K | 164K | — | +187% |
| Cortecs | $0.50 | $1.50 | $0.15 | 164K | 66K | — | +220% |
| Friendli | $0.50 | $1.50 | $0.25 | 164K | 147K | 100%24h | +220% |
| $0.56 | $1.68 | — | 164K | 66K | 36.9%24h | +259% | |
| Azure | $0.58 | $1.68 | — | 128K | 128K | — | +265% |
| Azure Cognitive Services | $0.58 | $1.68 | — | 128K | 128K | — | +265% |
| Vercel AI Gateway | $0.62 | $1.85 | — | 128K | 8K | — | +296% |
| Amazon Bedrock | $0.62 | $1.85 | — | 164K | 82K | — | +296% |
| Phala | $1.00 | $1.00 | $0.50 | 164K | 64K | 100%24h | +327% |
| Chutes | $1.00 | $1.00 | $0.10 | 131K | 66K | — | +327% |
| NEAR AI Cloud | $1.10 | $1.10 | $0.55 | 128K | 64K | — | +370% |
| SambaNovano tool calling | $3.00 | $4.50 | — | 33K | 7K | 95.7%24h | +1341% |
| Marano tool calling | $3.00 | $4.50 | — | 33K | 7K | 0%24h | +1341% |
Value
Price vs. performance
DeepSeek V3.2 Other DeepSeek Other labs Value frontier · Terminal-Bench vs. input price per 1M tokens (log scale) · Drag to pan, pinch or Ctrl + scroll to zoom, double-click to zoom in.
Benchmarks
11 benchmark scores
| Benchmark | Score | Rank | Leader |
|---|---|---|---|
| Terminal-BenchSoftware engineering · Tasks completed with the programs in a real terminal (best agent) | 39.6% | #23of 40 | GPT-5.584.7% |
| SciCodeSoftware engineering · Research-coding problems from the natural sciences | 38.9% | #79of 102 | Claude Opus 5.566.9% |
| Aider PolyglotSoftware engineering · Hard Exercism problems in many languages | 74.2% | #4of 26 | GPT-588.0% |
| APEX-AgentsAgents & computer use · Long professional tasks in banking, consulting and law | 21.3% | #49of 51 | Gemini 4 Argon82.4% |
| Vending-Bench 2Agents & computer use · Run a simulated vending business for a year (final balance) | $1,034 | #47of 61 | GPT-6 Astra$15,515 |
| Fiction.LiveBenchWriting & content · Understanding long creative writing (16k tokens) | 83.3% | #5of 24 | GPT-597.2% |
| WebDev ArenaDesign & visual · People vote on which model builds the better website (Elo) | 1367 | #72of 90 | Claude Opus 5.51820 |
| ARC-AGI-2Reasoning & knowledge · Few-shot abstract reasoning on grid puzzles | 4.0% | #61of 75 | GPT-6 Astra95.0% |
| ARC-AGI-1Reasoning & knowledge · Pattern induction from input–output examples | 57.0% | #55of 75 | Gemini 4 Argon98.5% |
| CritPtReasoning & knowledge · Open-ended, research-level physics problems | 2.9% | #65of 111 | GPT-5.6 Sol32.3% |
| LMCAReasoning & knowledge · Judge the quality of conceptual arguments like experts | 29.1% | #82of 123 | Claude Sonnet 5.568.3% |
Specs
Details
- Model ID
deepseek/deepseek-v3.2- Developer
- DeepSeek
- Release date
- Dec 1, 2025 ·
- Context window
- 163,840 tokens
- Max output
- 65,536 tokens
- API price
- $0.28 / $0.42 per 1M input / output tokens
- Providers
- 43
- Input types
- text
- Reasoning
- Yes
- Tool calling
- Yes
- Open weights
- Yes
- Free tier
- No
Release date and prices from OpenRouter's catalog; providers from OpenRouter and models.dev. Buying directly from a provider can cost a little differently. Speed, latency and uptime are measured on OpenRouter only; uptime is the share of OpenRouter requests each provider served over the last day. Benchmark scores are the best result per benchmark from Epoch AI, licensed CC BY 4.0. Rank counts every model tracked here with a score.