DeepSeek V3.1
30 cheaper ranked models score higher: DeepSeek V4.1 Flash (58.6, $0.518), DeepSeek V4.1 Flash 1017 (58.6, $0.525), GPT-5.6 Luna (58.3, $0.450) and 27 more.
As of 12 Nov 2026, DeepSeek V3.1 by DeepSeek scores 52.5 on the BenchLeader Index (603 ranked configurations), with its best results in reasoning (59) and weakest in instruction following (44). API pricing: $0.574 per 1M input tokens and $1.70 per 1M output tokens ($0.855 blended at 3:1), list price on 12 Nov 2026. It streams about 92 tokens/s (Artificial Analysis). Released 21 Aug 2025.
Key facts (as of 12 Nov 2026)
- DeepSeek V3.1 API pricing: $0.574 per 1M input tokens and $1.70 per 1M output tokens ($0.855 blended at 3:1) (list price, 12 Nov 2026).
- DeepSeek V3.1 streams about 92 tokens/s (Artificial Analysis, 12 Nov 2026).
- DeepSeek V3.1 was released on 21 Aug 2025 (DeepSeek).
In detail. DeepSeek V3.1 is a DeepSeek open-weights reasoning model, released 21 Aug 2025. Its best configuration (thinking reasoning effort) ranks #274 of 603 on the BenchLeader Index at 52.5 ±5.1, in the upper half. The ± is the point: 289 other configurations score within that range, so they and this one cannot be told apart on quality alone — price and speed are what separate them. It scores highest in reasoning (59) and lowest in instruction following (44). At $0.658 per million tokens blended it is mid-priced. Output speed of 92 tokens per second puts it faster than most, with a first answer in 10.4 s. It has been measured at 3 reasoning-effort settings; this summary describes the best-scoring one, and the tabs above switch between them. Last measured 9 Nov 2026.
- Blended price
- $0.658/M
- $0.430 in · $1.34 out
- via Artificial Analysis
- Output speed
- 92 tok/s
- measured by Artificial Analysis
- First answer
- 10 s
- first token 0.51 s
- Context
- 131k
- via models.dev
- Full answer
- 27 s
- median, reasoning included
- Cost per task
- –
- Tokens per task
- –
- Released
- 21 Aug 2025
- Overall index53
- Reasoning59
- Coding50
- Agents & tools56
- Maths52
- Instruction following44
- Human preference59
- Long context53
Embed this rank
More badge styles[](https://www.benchleader.com/models/deepseek-v3-1)Paste it at the top of README.md, beside your other badges. The badge redraws with every data update and links back to this page.
Versions
DeepSeek has shipped 2 models under this name. Each is ranked on its own results; a newer version often has fewer results so far, which holds its index nearer the average until more arrive.
| Model | Released | Index | Rank |
|---|---|---|---|
| DeepSeek V3.2thinking | 1 Dec 2025 | 56.3 | #187 |
| DeepSeek V3.1thinkingthis page | 21 Aug 2025 | 52.5 | #274 |
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. “Not stated” collects results from publishers that did not say which setting they used; for a reasoning model that is usually its thinking mode, but we do not assume it. Pick a setting here or at the top of the page to see its price, speed and category scores.
| Effort | Index | Rank | Speed | First answer | Chat reply cost | Categories |
|---|---|---|---|---|---|---|
| no reasoning | – | – | 70 tok/s | 2.99 s | $0.0007 | Agents & tools 55 · Instruction following 41 · Long context 48 |
| thinkingbest | 52.5 | #274 | 92 tok/s | 10 s | $0.0006 | Agents & tools 56 · Coding 50 · Human preference 59 · Instruction following 44 · Long context 53 · Maths 52 · Reasoning 59 |
| not stated | 47.9 | #390 | 32 tok/s | 0.51 s | $0.0007 | Coding 49 · Human preference 59 · Instruction following 34 · Knowledge 36 · Long context 48 · Maths 59 · Reasoning 48 |
How its index has moved
8 Sep 2026 to 12 Nov 2026- thinking
- not stated
The index is recomputed from scratch every day, so a line moves when a new benchmark result lands, when a publisher revises a score, or when the models it is normalised against change. Early movement usually means the score is still settling.
Benchmark results
See all LLM benchmarks. 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 | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| SimpleBencheffort not stated | – | – | 40.0%#79 | SimpleBench | |
| Arena Hard Prompts | – | 1437#134 | 1433#140 | Arena (LMArena) | |
| GPQA Diamond (AA)not in index | 73.5%#271 | 77.9%#219 | – | Artificial Analysis | |
| Humanity's Last Exam (AA)not in index | 6.7%#403 | 14.3%#274 | – | Artificial Analysis | |
| Kagi LLM Benchmarkeffort not stated | – | – | 53.2%#71 | Kagi LLM Benchmark | |
| CritPtnot in index | 0.0%#335 | 2.0%#218 | – | Artificial Analysis | |
| LMCAnot in index | – | 24.3%#159 | – | Epoch AI Benchmarking Hub | |
| DTBenchnot in index | – | 82.7%#125 | – | Epoch AI Benchmarking Hub | |
| ForecastBenchnot in indexeffort not stated | – | – | 60.5%#74 | ForecastBench |
Coding
| Benchmark | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| WeirdML | – | 38.4%#112 | 37.5%#118 | WeirdML | |
| Arena Coding | – | 1455#147 | 1448#155 | Arena (LMArena) | |
| Design Arena Websitenot in index | – | 1182#97 | 1117#134 | Design Arena | |
| LiveCodeBench (AA)not in index | 57.7%#117 | 78.4%#33 | – | Artificial Analysis |
Agents & tools
| Benchmark | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| Terminal-Bench Hard | 24.2%#147 | 25.0%#141 | – | Artificial Analysis | |
| τ²-Bench Telecom (AA)not in index | 34.8%#241 | 37.4%#230 | – | Artificial Analysis | |
| Poker Agentnot in indexeffort not stated | – | – | 1058.2%#7 | Vals AI | |
| Terminal-Bench 2.0 (Vals)not in indexeffort not stated | – | – | 22.5%#56 | Vals AI | |
| BrowseCompnot in indexeffort not stated | – | – | 2.6%#9 | OpenAI |
Maths
| Benchmark | thinking | not stated | Source | Trend |
|---|---|---|---|---|
| MathArena Apexpossible contamination | 0.5%#42 | – | MathArena | |
| Arena Maths | 1415#133 | 1413#135 | Arena (LMArena) |
Knowledge
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| CorpFin | 51.5%#101 | Vals AI | |
| PRBench Finance | 35.1%#37 | Scale AI SEAL | |
| PRBench Legal | 37.6%#41 | Scale AI SEAL | |
| MultiNRCpossible contamination | 23.6%#32 | Scale AI SEAL | |
| CaseLaw v2not in index | 53.9%#39 | Vals AI |
Instruction following
| Benchmark | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| IFBench | 37.8%#293 | 41.5%#249 | – | Artificial Analysis | |
| MultiChallengeeffort not stated | – | – | 46.1%#30 | Scale AI SEAL | |
| Arena Instruction Followingnot in index | – | 1416#118 | 1402#143 | Arena (LMArena) |
Human preference
| Benchmark | thinking | not stated | Source | Trend |
|---|---|---|---|---|
| Arena Text | 1416#136 | 1417#133 | Arena (LMArena) | |
| Arena Creative Writingnot in index | 1402#107 | 1388#131 | Arena (LMArena) | |
| Arena Multi-turnnot in index | 1414#140 | 1404#154 | Arena (LMArena) | |
| Arena Longer Queriesnot in index | 1446#99 | 1421#137 | Arena (LMArena) |
Long context
| Benchmark | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| Fiction.LiveBench 120keffort not stated | – | – | 53.1%#20 | Fiction.live | |
| AA-LCR v1.1 | 47.0%#348 | 56.7%#312 | – | Artificial Analysis |
Composite
| Benchmark | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| Epoch Capabilities Indexnot in indexeffort not stated | – | – | 140.8#112 | Epoch AI Benchmarking Hub | |
| AA Intelligence Index v4.3.2not in index | – | 13.5#317 | – | Artificial Analysis | |
| AA Progress Index v1not in index | 212.7#374 | 233.7#308 | – | Artificial Analysis |
Safety & honesty
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| FORTRESSnot in index | 60.5%#5 | Scale AI SEAL | |
| MASKnot in indexpossible contamination | 46.3%#61 | Scale AI SEAL |
API pricing: 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 |
|---|---|---|---|---|---|---|
| CoreWeave | 66 tok/s | 0.27 s | $0.500 | $1.50 | 161k | fp8 |
| DeepInfra | 28 tok/s | 0.55 s | $0.250 | $0.950 | 164k | fp4 |
| SiliconFlow | 18 tok/s | 2.25 s | $0.270 | $1.00 | 164k | fp8 |
Price history
The price above, per 1M tokens, every day it changed.
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 $0.130 per 1M. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0006 | $0.0005 | 13.6 s |
| Summarize a 30-page report | 12,000 / 600 | $0.0060 | $0.0033 | 16.9 s |
| Code edit | 6,000 / 1,500 | $0.0046 | $0.0032 | 26.6 s |
| Agentic coding session | 60,000 / 4,000 | $0.031 | $0.018 | 53.7 s |
| Structured extraction | 2,000 / 200 | $0.0011 | $0.0007 | 12.5 s |
More from DeepSeek
See also
Embed or cite
Show where DeepSeek V3.1 ranks on a model card or README. The badge redraws itself with every daily update, carries the data date and links back to this page.
[](https://www.benchleader.com/models/deepseek-v3-1)Cite this
The ranking changes daily, so a reference carries the data date. APA style:
BenchLeader. (2026, November 12). DeepSeek V3.1 on the BenchLeader Index (method v1.4) [Data set]. Retrieved November 12, 2026, from https://www.benchleader.com/models/deepseek-v3-1
BibTeX:
@misc{benchleader_deepseek_v3_1_20261112,
author = {{BenchLeader}},
title = {{DeepSeek V3.1 on the BenchLeader Index (method v1.4)}},
year = {2026},
month = nov,
howpublished = {\url{https://www.benchleader.com/models/deepseek-v3-1}},
note = {Data as of 2026-11-12}
}Every badge style, the embeddable top-10 widget and how to cite the Index
Questions people ask
- What are DeepSeek V3.1's benchmark scores?
- As of 12 Nov 2026, DeepSeek V3.1 (thinking effort) scores 52.5 ±5.1 on the BenchLeader Index, strongest in reasoning (59) and weakest in instruction following (44). That rests on independent benchmark results only, and every score on this page links to its publisher.
- How does DeepSeek V3.1 rank?
- DeepSeek V3.1 (thinking effort) is #274 of 603 ranked configurations on 12 Nov 2026. 289 other configurations score within its margin, so the exact position matters less than the band. The leader is Claude Fable 5.2 (max effort) at 75.0.
- How much does DeepSeek V3.1 cost per 1M tokens?
- As of 12 Nov 2026, DeepSeek V3.1 costs $0.430 per 1M input tokens and $1.34 per 1M output tokens, $0.658 blended (three input tokens per output token). The list price last changed on 6 Nov 2026.
- When was DeepSeek V3.1 released?
- DeepSeek released DeepSeek V3.1 on 21 Aug 2025. BenchLeader has results for it at 3 reasoning-effort settings.
Data as of 12 Nov 2026. Compare these configurations.
Cite as: BenchLeader, “DeepSeek V3.1: benchmarks, pricing, speed and rank”, data as of 12 Nov 2026, https://www.benchleader.com/models/deepseek-v3-1.