DeepSeekDeepSeek V3.1
Best configuration ranks #190 of 610 on the BenchLeader Index at 54.8 ±4.2 (thinking reasoning effort). Last measured 2 Sept 2026. Released 21 Aug 2025.
- Blended price
- $1.92/M
- $1.64 in · $2.75 out
- Output speed
- 29 tok/s
- First answer
- 1.81 s
- Context
- 128k
- Overall index55
- Reasoning61
- Coding51
- Agents & tools60
- Knowledge52
- Instruction following57
- Human preference60
- Long context61
- Composite48
Reasoning-effort configurations
The same model behaves differently depending on how much it is allowed to think. Each row is one setting, scored only on the benchmarks that were run at that setting. “Default” means the publisher did not say which setting was used.
| Effort | Index | Rank | Speed | First answer | Chat reply cost | Categories |
|---|---|---|---|---|---|---|
| thinkingbest | 54.8 | #190 | 29 tok/s | 1.81 s | $0.0015 | Agents & tools 60 · Coding 51 · Composite 48 · Human preference 60 · Instruction following 57 · Knowledge 52 · Long context 61 · Reasoning 61 |
| default | 51.0 | #272 | 29 tok/s | 1.81 s | $0.0004 | Agents & tools 61 · Coding 51 · Composite 47 · Human preference 60 · Instruction following 43 · Knowledge 43 · Long context 48 · Reasoning 56 |
Benchmark results
One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.
Reasoning
| Benchmark | thinking | default | Source | Trend |
|---|---|---|---|---|
| LMArena Hard Prompts | 1446#99 | 1423#124 | LMArena | |
| GPQA Diamond (AA)not in index | 79.2%#194 | 75.0%#243 | Artificial Analysis | |
| Humanity's Last Exam (AA)not in index | 16.4%#184 | 8.7%#279 | Artificial Analysis | |
| Kagi LLM Benchmark | – | 57.4%#59 | Kagi LLM Benchmark |
Coding
| Benchmark | thinking | default | Source | Trend |
|---|---|---|---|---|
| SciCode | – | 40.6%#104 | SciCode | |
| WeirdML | 38.4%#110 | – | WeirdML | |
| LMArena Coding | 1461#111 | 1440#137 | LMArena | |
| SciCode (AA)not in index | 38.0%#126 | – | Artificial Analysis |
Agents & tools
| Benchmark | thinking | default | Source | Trend |
|---|---|---|---|---|
| Terminal-Bench Hard | 30.3%#110 | 31.8%#99 | Artificial Analysis | |
| τ²-Bench Telecom (AA)not in index | 37.4%#220 | 37.1%#222 | Artificial Analysis |
Knowledge
| Benchmark | thinking | default | Source | Trend |
|---|---|---|---|---|
| AA-Omniscience | -26.4#202 | -43.5#276 | Artificial Analysis |
Instruction following
| Benchmark | thinking | default | Source | Trend |
|---|---|---|---|---|
| IFBench | 57.0%#137 | 41.2%#244 | Artificial Analysis |
Human preference
| Benchmark | thinking | default | Source | Trend |
|---|---|---|---|---|
| LMArena Text | 1418#109 | 1415#112 | LMArena |
Long context
| Benchmark | thinking | default | Source | Trend |
|---|---|---|---|---|
| AA-LCR | 69.3%#170 | 45.3%#268 | Artificial Analysis |
Composite
| Benchmark | thinking | default | Source | Trend |
|---|---|---|---|---|
| AA Intelligence Index | 15.4#217 | 13.9#236 | Artificial Analysis |
Where to run it
Every provider serving this model through OpenRouter, with throughput and first-token latency measured on live traffic over the last 30 minutes and each provider’s own price. Purple marks the best in each column.
| Provider | Speed | First token | Input $/M | Output $/M | Context | Quantisation |
|---|---|---|---|---|---|---|
| StreamLake | 46 tok/s | 1.70 s | $0.343 | $1.03 | 128k | – |
| AtlasCloud | 36 tok/s | 1.93 s | $0.300 | $0.950 | 131k | fp8 |
| NovitaAI | 21 tok/s | 2.21 s | $0.270 | $1.00 | 131k | fp8 |
| SiliconFlow | 18 tok/s | 1.38 s | $0.270 | $1.00 | 164k | fp8 |
Price history
Listed price per 1M tokens over time, as recorded by OpenRouter for the provider with the longest history.
What a task costs
Estimates from list price, output speed and time to first answer for the best configuration. “With caching” assumes three-quarters of the input is served from the prompt cache at $0.135 per 1M. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0015 | $0.0010 | 12.3 s |
| Summarise a 30-page report | 12,000 / 600 | $0.021 | $0.0078 | 22.9 s |
| Code edit | 6,000 / 1,500 | $0.014 | $0.0072 | 54.4 s |
| Agentic coding session | 60,000 / 4,000 | $0.109 | $0.042 | 2.4 min |
| Structured extraction | 2,000 / 200 | $0.0038 | $0.0016 | 8.8 s |
See also
Data as of 9 Sept 2026. Compare these configurations.