DeepSeekDeepSeek V4 Flash
DeepSeek V4 Flash is a DeepSeek open-weights reasoning model, released 31 Jul 2026. Its best configuration (max reasoning effort) ranks #101 of 775 on the BenchLeader Index at 61.1 ±3.8, in the upper half. The ± is the point: 190 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 instruction following (77) and lowest in knowledge (49). At $0.175 per million tokens blended it is among the cheapest fifth of ranked models. Output speed of 145 tokens per second puts it in the fastest quarter, with a first answer in 120.7 s. It has been measured at 5 reasoning-effort settings; this summary describes the best-scoring one, and the tabs above switch between them. Last measured 31 Jul 2026.
- Blended price
- $0.175/M
- $0.140 in · $0.280 out
- Output speed
- 145 tok/s
- measured by Artificial Analysis
- First answer
- 121 s
- first token 1.27 s
- Context
- 1.0M
- Full answer
- 131 s
- median, reasoning included
- Cost per run
- $0.174
- one full Intelligence Index run
- Released
- 31 Jul 2026
- Overall index61
- Reasoning64
- Coding56
- Agents & tools67
- Maths56
- Knowledge49
- Instruction following77
- Long context65
- Composite68
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 | 48.2 | #428 | 50 tok/s | 1.27 s | $0.0001 | Agents & tools 64 · Composite 51 · Instruction following 49 · Knowledge 42 · Long context 45 · Reasoning 35 |
| low | – | – | 50 tok/s | 1.27 s | $0.0002 | Reasoning 56 |
| high | 58.6 | #150 | 50 tok/s | 1.27 s | $0.0001 | Agents & tools 59 · Coding 56 · Composite 57 · Human preference 62 · Instruction following 72 · Knowledge 54 · Long context 61 · Maths 62 · Reasoning 55 |
| maxbest | 61.1 | #101 | 145 tok/s | 121 s | $0.0001 | Agents & tools 67 · Coding 56 · Composite 68 · Instruction following 77 · Knowledge 49 · Long context 65 · Maths 56 · Reasoning 64 |
| not stated | 54.4 | #247 | 50 tok/s | 1.27 s | $0.0002 | Agents & tools 54 · Coding 46 · Human preference 62 · Maths 59 · Reasoning 58 |
How its index has moved
8 Sept 2026 to 19 Oct 2026- max
- high
- not stated
- no reasoning
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 | low | high | max | not stated | Source | Trend |
|---|---|---|---|---|---|---|---|
| GPQA Diamond | – | – | – | 91.0%#32 | – | Epoch AI Benchmarking Hub | |
| SimpleBencheffort not stated | – | – | – | – | 61.1%#26 | SimpleBench | |
| Arena Hard Prompts | – | – | 1458#100 | – | 1459#96 | Arena (LMArena) | |
| LiveBench 1.0 Reasoningnot in indexeffort not stated | – | – | – | – | 86.6%#32 | LiveBench | |
| GPQA Diamond (AA)not in index | 71.6%#282 | – | 86.7%#105 | 90.8%#55 | – | Artificial Analysis | |
| Humanity's Last Exam (AA)not in index | 7.8%#352 | – | 30.3%#145 | 38.5%#96 | – | Artificial Analysis | |
| GPQA Diamond (Vals)not in index | – | – | 89.9%#27 | – | – | Vals AI | |
| Kagi LLM Benchmarkeffort not stated | – | – | – | – | 52.2%#76 | Kagi LLM Benchmark | |
| ARC-AGI-1 | 11.8%#219 | 84.0%#106 | 87.0%#91 | 89.0%#79 | – | ARC Prize | |
| ARC-AGI-2 | 2.1%#190 | 46.0%#105 | 56.0%#93 | 61.4%#75 | – | ARC Prize | |
| CritPt | 0.3%#266 | – | 3.4%#172 | 16.6%#84 | – | Artificial Analysis | |
| Chess Puzzlesnot in index | – | – | – | 32.0%#63 | – | Epoch AI Benchmarking Hub | |
| Mystery Game Puzzlesnot in index | – | – | – | 34.0%#24 | – | Epoch AI Benchmarking Hub | |
| LMCAnot in index | – | – | – | 41.7%#102 | – | Epoch AI Benchmarking Hub | |
| DTBenchnot in index | – | – | – | 90.9%#76 | – | Epoch AI Benchmarking Hub |
Coding
| Benchmark | no reasoning | high | max | not stated | Source | Trend |
|---|---|---|---|---|---|---|
| SciCode | – | 42.0%#151 | 49.9%#111 | – | SciCode | |
| WeirdML | – | 57.0%#56 | 63.0%#43 | – | WeirdML | |
| FrontierCodeeffort not stated | – | – | – | 18.8%#39 | Cognition | |
| Arena Coding | – | 1480#106 | – | 1484#100 | Arena (LMArena) | |
| Arena WebDev | – | 1582#34 | – | – | Arena (LMArena) | |
| LiveBench 1.0 Codingnot in indexeffort not stated | – | – | – | 75.0%#52 | LiveBench | |
| SciCode (AA)not in index | – | 40.2%#169 | 50.4%#105 | – | Artificial Analysis | |
| LiveCodeBench | – | 87.3%#16 | – | – | Vals AI | |
| IOI | – | 32.7%#39 | – | – | Vals AI | |
| SWE-bench (Vals)not in index | – | 88.8%#11 | – | – | Vals AI | |
| Code Migrationnot in index | – | 38.6%#32 | – | – | Vals AI | |
| ProgramBenchnot in index | – | 0.0%#32 | – | – | Vals AI | |
| Vibe Code Bench v1.1not in index | – | 74.7%#35 | – | – | Vals AI | |
| ALE-Benchnot in index | 325.0#116 | 678.2#87 | 1306.1#33 | – | Epoch AI Benchmarking Hub | |
| Terminal-Bench 4.0 (AA)not in index | – | 3.0%#106 | 12.1%#82 | – | Artificial Analysis | |
| Terminal-Bench 2.1 (AA)not in index | – | 56.9%#101 | 78.7%#49 | – | Artificial Analysis |
Agents & tools
Maths
| Benchmark | high | max | not stated | Source | Trend |
|---|---|---|---|---|---|
| FrontierMath Tiers 1–3 | – | 57.5%#43 | – | Epoch AI Benchmarking Hub | |
| FrontierMath Tier 4 | – | 24.4%#49 | – | Epoch AI Benchmarking Hub | |
| OTIS Mock AIME | – | 94.4%#53 | – | Epoch AI Benchmarking Hub | |
| ProofBencheffort not stated | – | – | 56.0%#24 | Vals AI | |
| LiveBench Mathematicsnot in indexeffort not stated | – | – | 86.8%#49 | LiveBench | |
| AIME 2026possible contamination | – | 95.8%#13 | – | MathArena | |
| HMMT February 2026possible contamination | – | 93.9%#10 | – | MathArena | |
| MathArena Apexpossible contamination | – | 27.1%#11 | – | MathArena | |
| Arena Maths | 1442#90 | – | 1425#113 | Arena (LMArena) |
Knowledge
| Benchmark | no reasoning | high | max | not stated | Source | Trend |
|---|---|---|---|---|---|---|
| SimpleQA Verified | – | – | 33.6%#57 | – | Epoch AI Benchmarking Hub | |
| LiveBench 1.0 Data Analysisnot in indexeffort not stated | – | – | – | 79.3%#14 | LiveBench | |
| AA-Omniscience | -44.0#338 | -23.1#253 | -14.3#215 | – | Artificial Analysis | |
| MMLU-Pro | – | 86.2%#45 | – | – | Vals AI | |
| LegalBench | – | 77.7%#104 | – | – | Vals AI | |
| CorpFin | – | 61.9%#52 | – | – | Vals AI | |
| TaxEval | – | 70.7%#84 | – | – | Vals AI | |
| BioMysteryBenchnot in index | – | 64.4%#21 | – | – | Vals AI | |
| Excel Modeling Benchmarknot in index | – | 57.0%#42 | – | – | Vals AI | |
| MedCodenot in index | – | 41.4%#52 | – | – | Vals AI | |
| MedScribenot in index | – | 80.4%#56 | – | – | Vals AI | |
| AA-Omniscience: accuracynot in index | 26.0%#217 | 34.9%#151 | 40.4%#120 | – | Artificial Analysis | |
| AA-Omniscience: non-hallucinationnot in index | 5.5%#514 | 11.0%#394 | 8.3%#461 | – | Artificial Analysis |
Instruction following
| Benchmark | no reasoning | high | max | not stated | Source | Trend |
|---|---|---|---|---|---|---|
| LiveBench Languagenot in indexeffort not stated | – | – | – | 79.2%#37 | LiveBench | |
| LiveBench Instruction Followingnot in indexeffort not stated | – | – | – | 65.5%#43 | LiveBench | |
| IFBench | 47.2%#188 | 73.5%#40 | 79.2%#11 | – | Artificial Analysis | |
| Arena Instruction Followingnot in index | – | 1434#89 | – | 1428#94 | Arena (LMArena) |
Human preference
| Benchmark | high | not stated | Source | Trend |
|---|---|---|---|---|
| Arena Text | 1439#95 | 1436#101 | Arena (LMArena) | |
| Arena Creative Writingnot in index | 1411#88 | 1408#92 | Arena (LMArena) | |
| Arena Multi-turnnot in index | 1447#95 | 1453#82 | Arena (LMArena) | |
| Arena Longer Queriesnot in index | 1449#90 | 1447#91 | Arena (LMArena) |
Long context
| Benchmark | no reasoning | high | max | Source | Trend |
|---|---|---|---|---|---|
| AA-LCR | 41.7%#347 | 72.0%#188 | 79.7%#87 | Artificial Analysis | |
| MLCRnot in index | – | 6.7%#71 | 13.3%#52 | Artificial Analysis |
Composite
| Benchmark | no reasoning | high | max | not stated | Source | Trend |
|---|---|---|---|---|---|---|
| Epoch Capabilities Indexnot in indexeffort not stated | – | – | – | 154.8#39 | Epoch AI Benchmarking Hub | |
| LiveBench 1.0not in indexeffort not stated | – | – | – | 74.2%#42 | LiveBench | |
| AA Intelligence Index v4.3.2 | 18.9#243 | 24.4#173 | 34.3#94 | – | 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 |
|---|---|---|---|---|---|---|
| Baidu Qianfan | 104 tok/s | 0.97 s | $0.140 | $0.280 | 1.0M | fp8 |
| Relace | 86 tok/s | 0.44 s | $0.050 | $0.450 | 1.0M | fp4 |
| Alibaba Cloud Int. | 81 tok/s | 1.18 s | $0.150 | $0.300 | 1M | fp8 |
| Cloudflare | 57 tok/s | 3.31 s | $0.440 | $1.32 | 384k | – |
| Azure | 53 tok/s | 1.92 s | $0.210 | $0.560 | 1.0M | – |
| GMICloud | 50 tok/s | 1.94 s | $0.091 | $0.182 | 1.0M | fp8 |
| SiliconFlow | 50 tok/s | 1.03 s | $0.130 | $0.280 | 1.0M | fp8 |
| Parasail | 47 tok/s | 0.90 s | $0.140 | $0.280 | 1.0M | fp8 |
| Wafer | 39 tok/s | 0.74 s | $0.700 | $0.170 | 1.0M | mxfp4 |
| StreamLake | 37 tok/s | 1.92 s | $0.083 | $0.165 | 1.0M | fp8 |
| AtlasCloud | 37 tok/s | 2.27 s | $0.140 | $0.280 | 1.0M | fp4 |
| DigitalOcean | 25 tok/s | 0.76 s | $0.098 | $0.196 | 1.0M | – |
| Mancer | 25 tok/s | 4.48 s | $0.190 | $0.500 | 1.0M | fp8 |
| Venice | 23 tok/s | 1.62 s | $0.097 | $0.193 | 1M | – |
| DeepInfra | 17 tok/s | 5.43 s | $0.090 | $0.180 | 1.0M | 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 $0.003 per 1M. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0001 | $0.0001 | 2.0 min |
| Summarize a 30-page report | 12,000 / 600 | $0.0018 | $0.0006 | 2.1 min |
| Code edit | 6,000 / 1,500 | $0.0013 | $0.0006 | 2.2 min |
| Agentic coding session | 60,000 / 4,000 | $0.0095 | $0.0034 | 2.5 min |
| Structured extraction | 2,000 / 200 | $0.0003 | $0.0001 | 2.0 min |
More from DeepSeek
Compare DeepSeek V4 Flash with…
- DeepSeek V4 Flash vs Claude Opus 5.5
- DeepSeek V4 Flash vs Gemini 4 Argon
- DeepSeek V4 Flash vs Muse Wonder Full
- DeepSeek V4 Flash vs GPT-6 Astra
- DeepSeek V4 Flash vs Claude Fable 5.1
- DeepSeek V4 Flash vs GPT-6.1 Sol
- DeepSeek V4 Flash vs Claude Sonnet 5.5
- DeepSeek V4 Flash vs Claude Fable 5
- DeepSeek V4 Flash vs Claude Opus 5
- DeepSeek V4 Flash vs GPT-5.6 Sol
See also
Data as of 21 Oct 2026. Compare these configurations.
Cite as: BenchLeader, “DeepSeek V4 Flash: benchmarks, pricing, speed and rank”, https://www.benchleader.com/models/deepseek-v4-flash, data as of 21 Oct 2026.