DeepSeekDeepseek v4 Flash Vision
Best configuration ranks #140 of 610 on the BenchLeader Index at 57.4 ±5.9. Last measured 25 Jun 2026. Released 21 Aug 2026.
- Blended price
- $0.175/M
- $0.140 in · $0.280 out
- Output speed
- 119 tok/s
- First answer
- 18 s
- first token 0.95 s
- Context
- 1M
- Overall index57
- Knowledge56
- Multimodal59
- Long context67
- Composite65
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 | default | Source |
|---|---|---|
| LiveBench Reasoningnot in index | 85.4%#28 | LiveBench |
| GPQA Diamond (AA)not in index | 91.3%#48 | Artificial Analysis |
| Humanity's Last Exam (AA)not in index | 34.5%#83 | Artificial Analysis |
Coding
| Benchmark | default | Source |
|---|---|---|
| LiveBench Codingnot in index | 68.2%#50 | LiveBench |
| SciCode (AA)not in index | 49.6%#72 | Artificial Analysis |
Agents & tools
| Benchmark | default | Source |
|---|---|---|
| LiveBench Agentic Codingnot in index | 65.1%#3 | LiveBench |
Maths
| Benchmark | default | Source |
|---|---|---|
| LiveBench Mathematicsnot in index | 87.8%#32 | LiveBench |
Knowledge
| Benchmark | default | Source |
|---|---|---|
| LiveBench Data Analysisnot in index | 79.5%#7 | LiveBench |
| AA-Omniscience | -17.6#173 | Artificial Analysis |
Instruction following
| Benchmark | default | Source |
|---|---|---|
| LiveBench Languagenot in index | 80.4%#21 | LiveBench |
Multimodal
| Benchmark | default | Source |
|---|---|---|
| MMMU-Pro | 74.8%#80 | Artificial Analysis |
Long context
| Benchmark | default | Source |
|---|---|---|
| AA-LCR | 81.3%#34 | Artificial Analysis |
Composite
| Benchmark | default | Source |
|---|---|---|
| LiveBench | 76.8%#19 | LiveBench |
| AA Intelligence Index | 35.0#53 | 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 |
|---|---|---|---|---|---|---|
| Fireworks | 114 tok/s | 0.94 s | $0.220 | $0.660 | 1.0M | – |
| AtlasCloud | 96 tok/s | 0.89 s | $0.440 | $1.32 | 1.0M | fp8 |
| DeepSeek | 94 tok/s | 0.95 s | $0.220 | $0.660 | 1.0M | – |
| DeepInfra | 80 tok/s | 1.43 s | $0.216 | $0.647 | 1.0M | fp8 |
| NovitaAI | 59 tok/s | 1.22 s | $0.440 | $1.32 | 1.0M | – |
| SiliconFlow | 52 tok/s | 0.80 s | $0.440 | $1.32 | 1.0M | fp8 |
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 | 20.3 s |
| Summarise a 30-page report | 12,000 / 600 | $0.0018 | $0.0006 | 22.8 s |
| Code edit | 6,000 / 1,500 | $0.0013 | $0.0006 | 30.4 s |
| Agentic coding session | 60,000 / 4,000 | $0.0095 | $0.0033 | 51.5 s |
| Structured extraction | 2,000 / 200 | $0.0003 | $0.0001 | 19.5 s |
See also
Data as of 9 Sept 2026. Compare with another model.