DeepSeek V4 Flash 0423
Fast DeepSeek V4 lane for economical reasoning, coding, and long-context work
- Context
- 1.0M
- Released
- 5mo ago
- Parameters
- —
- Input / 1M tokens
- $0.09
- Output / 1M tokens
- $0.18
- Cached input / 1M tokens
- $0.02
Sources:models.devOpenRouter
Benchmarks
DeepSWE is the primary coding benchmark where an exact model evaluation exists. Other sourced evaluations appear below; missing scores are not zero.
DeepSWE 1.1
Primary coding-agent benchmark · 113 tasks · Data generated 2026-09-22
- pass@4
- 80.5%
- Average cost / attempt
- $0.10
- Benchmark runs
- 4
pass@1 is the success rate across scored attempts; pass@4 is the share of tasks solved in at least one of four attempts. Costs are reported by DeepSWE for each configuration. These results are not directly comparable to evaluations using different harnesses. Official source JSON.
Specifications
- Developer
- DeepSeek
- Model family
- DeepSeek
- Type
- Language model
- Predecessor
- —
- Successor
- —
- Released
- April 24, 2026
- Model ID
deepseek/deepseek-v4-flash- Openness
- Open weights
- License
- —
- Parameters
- —
- Input modalities
- text
- Output modalities
- text
- Modality
- text->text
- Tokenizer
- DeepSeek
- Max output tokens
- 944K
- Knowledge cutoff
- 2025-05
- Hugging Face
- deepseek-ai/DeepSeek-V4-Flash
- Source freshness
- Metadata synced Oct 8, 2026, 10:17 PM UTC
Additional source evidence
- Source model
- deepseek/deepseek-v4-flash
- Source family
- deepseek-flash
- Last updated
- 2026-04-24
- Knowledge cutoff
- 2025-05
- Model context
- 1M
- Model output limit
- 384K
- Open weights
- Yes
- File attachments
- No
- Reasoning
- Yes
- Tool calling
- Yes
- Structured output
- Yes
- Temperature control
- Yes
Downloadable weights
Published benchmarks
| Benchmark | Score | Evaluation details | Evidence |
|---|---|---|---|
| SWE-Bench Verified | 79 · resolved | — | Source |
| MMLU-Pro | 86.2 · EM | preview checkpoint; max effort | Source |
| SimpleQA-Verified | 34.1 · pass@1 | preview checkpoint; max effort | Source |
| Chinese SimpleQA | 78.9 · pass@1 | preview checkpoint; max effort | Source |
| GPQA Diamond | 88.1 · pass@1 | preview checkpoint; max effort | Source |
| Humanity's Last Exam | 34.8 · pass@1 | preview checkpoint; max effort; without tools | Source |
| LiveCodeBench | 91.6 · pass@1 | preview checkpoint; max effort | Source |
| Codeforces | 3052 · rating | preview checkpoint; max effort | Source |
| HMMT | 94.8 · pass@1 | February 2026 · preview checkpoint; max effort | Source |
| IMOAnswerBench | 88.4 · pass@1 | preview checkpoint; max effort | Source |
| MathArena Apex | 33 · pass@1 | preview checkpoint; max effort | Source |
| MathArena Apex Shortlist | 85.7 · pass@1 | preview checkpoint; max effort | Source |
| MRCR | 78.7 · MMR | 1M context · preview checkpoint; max effort | Source |
| CorpusQA | 60.5 · accuracy | 1M context · preview checkpoint; max effort | Source |
| Terminal-Bench | 56.9 · accuracy | preview checkpoint; max effort · 2.0 | Source |
| SWE-Bench Pro | 52.6 · resolved | preview checkpoint; max effort | Source |
| SWE-Bench Multilingual | 73.3 · resolved | preview checkpoint; max effort | Source |
| BrowseComp | 73.2 · pass@1 | preview checkpoint; max effort | Source |
| Humanity's Last Exam | 45.1 · pass@1 | preview checkpoint; max effort; with tools | Source |
| MCP Atlas | 69 · pass@1 | public · preview checkpoint; max effort | Source |
| GDPval-AA | 1395 · Elo | preview checkpoint; max effort | Source |
| Toolathlon | 47.8 · pass@1 | preview checkpoint; max effort | Source |
Complete source record (JSON)
{
"id": "deepseek/deepseek-v4-flash",
"name": "DeepSeek V4 Flash",
"description": "Fast DeepSeek V4 lane for economical reasoning, coding, and long-context work",
"family": "deepseek-flash",
"last_updated": "2026-04-24",
"knowledge": "2025-05",
"attachment": false,
"reasoning": true,
"tool_call": true,
"structured_output": true,
"temperature": true,
"modalities": {
"input": [
"text"
],
"output": [
"text"
]
},
"limit": {
"context": 1000000,
"output": 384000
},
"benchmarks": [
{
"name": "SWE-Bench Verified",
"score": 79,
"source": "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash",
"metric": "resolved"
},
{
"name": "MMLU-Pro",
"score": 86.2,
"source": "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash",
"metric": "EM",
"variant": "preview checkpoint; max effort"
},
{
"name": "SimpleQA-Verified",
"score": 34.1,
"source": "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash",
"metric": "pass@1",
"variant": "preview checkpoint; max effort"
},
{
"name": "Chinese SimpleQA",
"score": 78.9,
"source": "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash",
"metric": "pass@1",
"variant": "preview checkpoint; max effort"
},
{
"name": "GPQA Diamond",
"score": 88.1,
"source": "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash",
"metric": "pass@1",
"variant": "preview checkpoint; max effort"
},
{
"name": "Humanity's Last Exam",
"score": 34.8,
"source": "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash",
"metric": "pass@1",
"variant": "preview checkpoint; max effort; without tools"
},
{
"name": "LiveCodeBench",
"score": 91.6,
"source": "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash",
"metric": "pass@1",
"variant": "preview checkpoint; max effort"
},
{
"name": "Codeforces",
"score": 3052,
"source": "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash",
"metric": "rating",
"variant": "preview checkpoint; max effort"
},
{
"name": "HMMT",
"score": 94.8,
"source": "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash",
"metric": "pass@1",
"variant": "preview checkpoint; max effort",
"dataset": "February 2026"
},
{
"name": "IMOAnswerBench",
"score": 88.4,
"source": "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash",
"metric": "pass@1",
"variant": "preview checkpoint; max effort"
},
{
"name": "MathArena Apex",
"score": 33,
"source": "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash",
"metric": "pass@1",
"variant": "preview checkpoint; max effort"
},
{
"name": "MathArena Apex Shortlist",
"score": 85.7,
"source": "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash",
"metric": "pass@1",
"variant": "preview checkpoint; max effort"
},
{
"name": "MRCR",
"score": 78.7,
"source": "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash",
"metric": "MMR",
"variant": "preview checkpoint; max effort",
"dataset": "1M context"
},
{
"name": "CorpusQA",
"score": 60.5,
"source": "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash",
"metric": "accuracy",
"variant": "preview checkpoint; max effort",
"dataset": "1M context"
},
{
"name": "Terminal-Bench",
"score": 56.9,
"source": "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash",
"metric": "accuracy",
"variant": "preview checkpoint; max effort",
"version": "2.0"
},
{
"name": "SWE-Bench Pro",
"score": 52.6,
"source": "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash",
"metric": "resolved",
"variant": "preview checkpoint; max effort"
},
{
"name": "SWE-Bench Multilingual",
"score": 73.3,
"source": "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash",
"metric": "resolved",
"variant": "preview checkpoint; max effort"
},
{
"name": "BrowseComp",
"score": 73.2,
"source": "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash",
"metric": "pass@1",
"variant": "preview checkpoint; max effort"
},
{
"name": "Humanity's Last Exam",
"score": 45.1,
"source": "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash",
"metric": "pass@1",
"variant": "preview checkpoint; max effort; with tools"
},
{
"name": "MCP Atlas",
"score": 69,
"source": "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash",
"metric": "pass@1",
"variant": "preview checkpoint; max effort",
"dataset": "public"
},
{
"name": "GDPval-AA",
"score": 1395,
"source": "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash",
"metric": "Elo",
"variant": "preview checkpoint; max effort"
},
{
"name": "Toolathlon",
"score": 47.8,
"source": "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash",
"metric": "pass@1",
"variant": "preview checkpoint; max effort"
}
],
"release_date": "2026-04-24",
"open_weights": true,
"weights": [
{
"label": "Hugging Face",
"url": "https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash"
}
]
}Infrastructure
Provider availability
| Provider | Endpoint | Context | Input /M tokens | Output /M tokens | Cached input /M tokens | Cache write /M tokens | Uptime 1d | Data handling |
|---|---|---|---|---|---|---|---|---|
| open-inference/fp4 | 1.0M | $0.05 | $1.15 | $0.01 | — | 95.60% | ZDR: YesTraining: Unknown Prompt logging: Unknown Retention: Unknown Checked 2026-10-15 | |
| relace/fp4 | 1.0M | $0.05 | $1.28 | $0.05 | — | 100.00% | ZDR: YesTraining: Unknown Prompt logging: Unknown Retention: Unknown Checked 2026-10-15 | |
| streamlake/fp8 | 1.0M | $0.08 | $0.17 | $0.02 | — | 100.00% | ZDR: NoTraining: Unknown Prompt logging: Unknown Retention: Unknown Checked 2026-10-15 | |
| deepinfra/fp8 | 1.0M | $0.09 | $0.18 | $0.02 | — | 99.25% | ZDR: YesTraining: Unknown Prompt logging: Unknown Retention: Unknown Checked 2026-10-15 | |
| gmicloud/fp8 | 1.0M | $0.09 | $0.18 | $0.02 | — | 100.00% | ZDR: NoTraining: Unknown Prompt logging: Unknown Retention: Unknown Checked 2026-10-15 | |
| venice | 1M | $0.10 | $0.19 | $0.02 | — | 100.00% | ZDR: YesTraining: Unknown Prompt logging: Unknown Retention: Unknown Checked 2026-10-15 | |
| digitalocean | 1.0M | $0.10 | $0.20 | $0.02 | — | 100.00% | ZDR: YesTraining: Unknown Prompt logging: Unknown Retention: Unknown Checked 2026-10-15 | |
| siliconflow/fp8 | 1.0M | $0.13 | $0.28 | $0.03 | — | 99.65% | ZDR: YesTraining: Unknown Prompt logging: Unknown Retention: Unknown Checked 2026-10-15 | |
| atlas-cloud/fp4 | 1.0M | $0.14 | $0.28 | $0.03 | — | 100.00% | ZDR: NoTraining: Unknown Prompt logging: Unknown Retention: Unknown Checked 2026-10-15 | |
| baidu/fp8 | 1.0M | $0.14 | $0.28 | $0.03 | — | 99.86% | ZDR: NoTraining: Unknown Prompt logging: Unknown Retention: Unknown Checked 2026-10-15 | |
| parasail/fp8 | 1.0M | $0.14 | $0.28 | $0.07 | — | 100.00% | ZDR: YesTraining: Unknown Prompt logging: Unknown Retention: Unknown Checked 2026-10-15 | |
| alibaba/fp8/us-east-1 | 1M | $0.15 | $0.30 | $0.03 | — | 99.93% | ZDR: NoTraining: Unknown Prompt logging: Unknown Retention: Unknown Checked 2026-10-15 | |
| mancer/fp8 | 1.0M | $0.19 | $0.50 | — | — | 97.35% | ZDR: YesTraining: Unknown Prompt logging: Unknown Retention: Unknown Checked 2026-10-15 | |
| azure/us | 1.0M | $0.21 | $0.56 | $0.03 | — | 97.36% | ZDR: YesTraining: Unknown Prompt logging: Unknown Retention: Unknown Checked 2026-10-15 | |
| cloudflare | 384K | $0.44 | $1.32 | $0.01 | — | 96.87% | ZDR: NoTraining: Unknown Prompt logging: Unknown Retention: Unknown Checked 2026-10-15 | |
| wafer | 1.0M | $0.48 | $0.17 | $0.01 | — | 99.79% | ZDR: YesTraining: Unknown Prompt logging: Unknown Retention: Unknown Checked 2026-10-15 |
Provider data:OpenRouter
Deployment graph
Can I run this?
Vertical Tabs only shows compatibility when it has structured evidence tying this exact model to hardware, runtime, quantization, memory use, and observed performance.