V4 Pro 0813
DeepSeek · released Aug 12, 2026
DeepSeek V4 Pro 0813 is a large-scale mixture-of-experts model from DeepSeek. This is the GA release of DeepSeek V4 Pro.
Specification
- Context window
- 1.05M
- Max output
- 393K
- Knowledge cutoff
- Not stated
- Parameters
- 1650.5B
- Licence
- mit
- Serving providers
- 20
- Moderated
- No
- Uptime
- 100.0%
Intelligence
36.0
88th percentile
Coding
68.8
Coding Index
Agentic
26.3
Agentic Index
Output speed
75 t/s
Median across providers
Latency
409ms
Time to first token
Cost per task
$0.38
Estimated
Benchmarks
Where the score comes from
The Intelligence Index is a composite. These are the underlying evaluations this model was actually measured on.
Evaluation scores
Percentage correct · higher is better
- GPQA Diamond92.8%
- AA-LCR (long context)80.3%
- SciCode51.0%
- Humanity's Last Exam41.0%
- τ²-bench (Telecom)39.6%
An evaluation missing from this list was not run for this model — it is not a zero.
View as table
| Evaluation | Score |
|---|---|
| GPQA Diamond | 92.8% |
| AA-LCR (long context) | 80.3% |
| SciCode | 51.0% |
| Humanity's Last Exam | 41.0% |
| τ²-bench (Telecom) | 39.6% |
Against its peers
Intelligence Index · this model highlighted, nearest peers in grey
- GPT-6 LunaOpenAI38.1
- MiMo-V2.6-FlashXiaomi37.9
- GPT-5.6 LunaOpenAI37.3
- V4 Pro 0813DeepSeek36.0
- V4 Pro 0423DeepSeek36.0
- V4 Flash 0731DeepSeek34.3
- V4 Flash 0423DeepSeek34.3
- Gemini 3.6 FlashGoogle34.0
Peers are the models sitting closest on the Intelligence Index — the set you would realistically choose between.
View as table
| Model | Intelligence |
|---|---|
| GPT-6 Luna | 38.1 |
| MiMo-V2.6-Flash | 37.9 |
| GPT-5.6 Luna | 37.3 |
| V4 Pro 0813 | 36.0 |
| V4 Pro 0423 | 36.0 |
| V4 Flash 0731 | 34.3 |
| V4 Flash 0423 | 34.3 |
| Gemini 3.6 Flash | 34.0 |
Percentile among all indexed models
Pricing
What it costs to run
List prices per million tokens, plus what one representative task works out to.
List price
- Input / 1M tokens
- $1.32
- Output / 1M tokens
- $3.96
- Cached input / 1M
- $0.044
- Blended 3:1
- $1.98
One task, estimated
$0.38
- Input tokens
- 50,000
- Output tokens
- 80,000
- Profile
- Reasoning
Estimated from list pricing: 50K input tokens plus 80K output tokens for reasoning models (25K for non-reasoning).
Serving providers
- Relace
- Wafer
- Ionstream
- StreamLake
- Sail Research
- Novita
- GMICloud
- Alibaba
- DeepInfra
- CoreWeave
- AtlasCloud
- Baidu
- Parasail
- Together
- DigitalOcean
- SiliconFlow
- Cloudflare
- DeepSeek
- Phala
- Venice
Speed and latency figures are medians across these providers, so a widely-served model reports a blend rather than any single endpoint.
Arena
Head-to-head generation quality
Elo from pairwise judgements, broken out by the kind of thing the model was asked to build.
- Overall Elo
- 1,254
- Win rate
- 56.8%
- Strongest at
- 3D scenes
- Tournaments
- 4,510
Elo by category
Dot position on a 1,125–1,275 scale · Elo has no meaningful zero
- 1,254
- 1,236
- 1,232
- 1,223
- 1,203
- 1,138
Agent categories (full-stack apps, mobile apps) are only scored for models tested in agent harnesses.
View as table
| Category | Elo | Win rate |
|---|---|---|
| 3D scenes | 1254 | 56.8% |
| Websites | 1236 | 50.5% |
| Game development | 1232 | 52.3% |
| UI components | 1223 | 50.6% |
| Data visualisation | 1203 | 48.7% |
| SVG | 1138 | 45.4% |